Where the science of drying medicinal cannabis stands, and what it means for supplier qualification

The water content of the fresh inflorescence is the most reproducible quantity in this literature. Every source whose text can be read puts it between 74.9 and 80.6 per cent, four titles from four independent groups, one of them a review that takes its figures from third parties. Both ends of that span come from a single study, and the four sources do not report the same quantity on the same scale, which is where the consensus comes apart. Baek and colleagues declare the basis for every figure and publish both formulas, which makes them the reference case of the corpus. Initial moisture runs 74.9 to 75.0 per cent (Hempress) and 77.4 to 80.6 per cent (Wild Bourbon), wet basis, falling to 1.9 to 6.8 per cent wet basis after drying (single study, one group, two cultivars, three plants per cultivar, experiments in triplicate) [9]. The 77 per cent often quoted from that work is a rounding that appears in the abstract with no counterpart in Table 1 [9]. Uziel and colleagues report 77.03 ± 0.75 per cent for the fresh inflorescence of the THC-dominant chemovar Infiniti and 13.10 ± 0.21 per cent after ten days of traditional drying, and the readable text states neither how they determined the moisture nor on which basis they express it (single study, one group, n=3) [80]. Placing that value on Baek’s scale requires an assumption. Das and colleagues write that “The inflorescence of raw cannabis contains around 78 to 80% moisture” with no basis attached, while Table 1 of the same review carries “Initial: 75-78% (wet basis, wb)”, a figure the review takes from third parties (review, no primary measurement) [31]. Yoon and colleagues open their cold-air series at 78.1 per cent relative water content, a different measurand (single study, one group, one cultivar) [83]. The work closest to industrial practice cannot be read at all. Crossref, OpenAlex and Semantic Scholar hold no deposited abstract for the industrial-scale tray drying study and Unpaywall reports it closed with no open location, so the work can be cited by title and by nothing else [32]. None of those figures comes from the determination the European monograph prescribes. Monograph 3028 sets loss on drying (2.2.32) at “maximum 12.0 per cent, determined on 1.000 g of the cut or milled herbal drug (not sieved) by drying over about 100 g of molecular sieve R at a pressure between 1.5 kPa and 2.5 kPa at 40 °C for 24 h”, and a loss weighed against the mass of drug taken for the test is a wet basis figure [22]. Baek determines moisture in an oven at 105 °C [9]. Birenboim and colleagues stop at constant weight, holding the weight change inside the error of a 10 ± 1 per cent water content, and the paper does not say which basis that figure uses (single study, one group, two chemovars, n=5) [10]. Punja and colleagues report that “the moisture content was estimated using the method following U.S. Pharmacopeia USP 731” (Methods, §2.8), which leaves the basis undeclared and deducible only from the method, since a loss on drying is expressed on the sample as taken [65]. Uziel declares neither method nor basis [80]. USP explains why the temperature matters, choosing vacuum at 40 °C for 24 hours “to avoid potential loss and other degradation not related to loss of water (e.g., from decarboxylation of acidic cannabinoids and loss of terpenes)” (Proposed For Comment Version 0.1; hmc.usp.org returned 403 and the wording reaches us through a secondary source) [79]. Ph. Eur. 3028 lands on the same combination of low temperature and vacuum [22]. An oven determination at 105 °C weighs water together with whatever the compendial conditions were written to protect, and no study in this corpus publishes a bridge between the reading of its own instrument and the compendial figure. Total water and free water are different quantities, and the European monograph specifies only the first. An exhaustive search of monograph 3028 returns no water activity specification [22]. Within this corpus, one study publishes measured water activity values on cannabis next to the matching moisture figures (single study, one group, n=3, two conditions), 13.10 ± 0.21 per cent moisture with aw 0.495 ± 0.002 after ten days at 16 °C and 50 per cent relative humidity, and 16.99 ± 0.42 per cent with aw 0.571 ± 0.002 after solid-state microwave drying at 50 °C for one hour and forty minutes [80]. Two pairs of values are not an isotherm. Punja and colleagues declare in their Methods that they measured the water activity of every sample following USP <922>, and they publish no value per sample; the aw figures in their text are a process target taken from Caplan and an approximation the authors mark as such, “translating to aw of ca. 0.62-0.7” for a moisture content of 12 to 14 per cent (single study, one group, twelve sampling positions per day, three repetitions) [65]. The published sorption isotherms for hemp floral material are on flower and leaf composite powders, and milling destroys the structure that governs internal moisture transport in the intact bud [59]. The only isotherm work on the whole inflorescence is a master’s thesis under embargo until 10 March 2027, whose repository record shows not even an abstract [81]. Converting a moisture figure into a water activity on this matrix, or the reverse, rests on no validated relationship. A moisture figure without its basis and its determination method cannot be compared with another one, and in this literature most figures arrive missing one or the other. Every moisture value quoted in the rest of this review carries both, or carries the statement that the authors gave neither.

The acid to neutral conversion begins before anyone applies heat. Kim and co-workers sampled the contents of the glandular trichome secretory cavities with glass microcapillaries, before any processing, and compared them with an extract of the same inflorescence after fifteen days of air drying in darkness at room temperature (single study, one group, one cultivar, cv. 'Cherry Wine', CBD dominant at 7.68% total CBD and 0.25% THC, one time zero against one fifteen day endpoint, replicate count not established). From the Results, “The ratio of CBD to CBDA was 1:99 (w/w) in secretory cavity contents, but 1:20 (w/w) in the inflorescence sample”, and “all acidic forms of cannabinoids have been reduced ranging from 0.5% to 2.4% to the corresponding neutral form” [48]. Birenboim and co-workers record the same direction on medicinal cannabis, with the chambers and the open air reference both set to 15 °C (single study, one group, one THCA dominant chemovar and one hybrid, n=5 per treatment). In chemovar '240' THCA falls by 2 to 20% across the four drying conditions, the smallest reduction under nitrogen and under controlled atmosphere and neither of those significant against time zero, while neutral THC in the open air reference reaches about 3.5 times its value at time zero. In 'Gen12' THCA and CBDA hold, with relative changes under 10% and none of them significant, CBGA falls between 11% under controlled atmosphere and 37% in open air, and neutral THC and CBD in open air reach roughly eight and three times their starting values [10]. Filer, in a narrative review, places the reaction earlier still, in the living plant, at about 0.5 mg/g of trichome weight for THC and CBD at week seven of flowering by quantitative NMR, and records an artefact that bears on older data, since a heated GC injector decarboxylates the acids during the analysis itself [38]. The two groups worked on different chemotypes and on different matrices. Drying changes the chemical identity of the material at temperatures that sit below any deliberate heating step. Heat makes the same reaction faster, and the effect follows the dose. Three research groups report it in four papers, and Chang Chen signs both [20] and [9]. Chen and co-workers take cannabidiolic acid conversion from 0.2% in ambient air to 14.1% at 90 °C, a conversion rate and not a content by weight (single study, three industrial hemp varieties, inflorescences mixed with leaves; abstract, introduction, conclusion and figures are public while Methods and Tables stay closed, so the replicate count is not verifiable). Their material is CBD dominant, and the transfer to THC dominant medicinal flos is our inference [20]. Hot air at 75 °C for 8 h gives 9.0 ± 3.0% decarboxylation on cv. Hempress and 8.5 ± 1.0% on cv. Wild Bourbon, against neutral cannabinoids below the detection limit in ambient air and in freeze drying (single study, two cultivars, three plants per cultivar, 200 g samples ground to 1 mm) [9]. The largest fall of the acidic precursors, up to threefold, comes with convective drying, while CBD and delta-9-THC rise up to 10 times under traditional drying on a dry matter basis, which is the authors' own figure and carries part of that rise as concentration on the shrinking dry denominator rather than as net gain (single study, one cultivar, n=3) [47]. Microwave with infrared takes THCA from 20.15 to 7.57% g/g dry solids while THC rises from 6.31 to 16.65% (single study; cultivar and replicates not declared in the abstract) [28]. The corpus holds no rate constant that applies to any of this. Its one kinetic data set comes from a dried extract film in a vial, under vacuum and in the dark, starting at 80 °C (single study, one extract preparation alongside pure standards). Wang and co-workers report first order kinetics, verified at 80 and 95 °C over the whole time scale and at 110 and 130 °C up to complete conversion, with k × 10³ in s⁻¹ of 0.18, 0.66 and 1.83 for THCA-A at 80, 95 and 110 °C, and activation energies of 88, 112 and 109 kJ/mol for THCA-A, CBDA and CBGA [82]. They write that “The rate constants for THCA-A were always approximately twice those of CBDA or CBGA” (Results), while their own constants give a THCA-A to CBDA ratio of 3.6 at 80 °C and 2.2 at 110 °C. The gap between the two acids is widest at 80 °C, the end of the measured range closest to drying. Below 80 °C no constant exists. Moreno and co-workers cover 80 to 160 °C on hemp plant material, with no activation energy citable from the publisher abstract (abstract only) [57]. Perrotin-Brunel and co-workers report pseudo first order kinetics with a barrier of 85 kJ/mol, while matrix, thermal range and replication sit behind a paywall, so 85 and 88 kJ/mol do not confirm each other [61]. Birenboim and co-workers set their chambers at 15 °C, inside what they report as the commonly used drying range of 15 to 21 °C [10], which leaves an extrapolation of 59 to 65 degrees below the coldest temperature at which anyone in this corpus measured a constant, and no study here validates it. Wang and co-workers also measured how far an extrapolation across matrices misses. At 110 °C, CBDA as a pure standard gives k = 0.16 × 10⁻³ s⁻¹ against 0.83 × 10⁻³ s⁻¹ for the same CBDA in the extract, five times slower, with a relative loss of 14% against 18% in the text [82]. Both matrices are simpler than an intact flower, where the CBD to CBDA ratio inside the secretory cavities is still 1:99 by weight [48]. A factor of five between a pure standard and a dried extract film, inside a single study, is a measured quantity, so constants published for one matrix do not predict another. The consequence lands on the release specification. Total cannabinoid content does not record the conversion. Baek and co-workers report that “The drying methods did not significantly influence the total cannabinoid contents (p > 0.01)” (Results) in the experiment where the post-curing acid to neutral ratio of cv. Hempress runs at 7.5 and 7.9 after hot air against 50.1 and 60.3 after freeze drying (Table 2) [9], and Das and co-workers report no significant effect on total THC (p ≥ 0.05) while THCA drops from 20.15 to 7.57% g/g dry solids [28]. Ph. Eur. 3028 sets the content requirement on total THC and total CBD, within ± 10 per cent of the declared label value, for herbal drug intended for prescription as a medicine [22]. A specification written on the total accepts the material at 7.5 and the material at 60.3 without distinguishing them. The acid to neutral ratio costs one number on the certificate, and with it you can tell the two apart.

Researchers report two quantities and readers treat them as one. A concentration on dry weight measures how concentrated the finished product is, while a retention figure against an unheated reference measures how much of the starting material survived. Drying removes most of the mass, between 75.8 ± 1.5% and 80 ± 4% of the fresh weight across the two chemovars and the two processes of Table 3 in [10] (the drying study, Plants 13(7):1049; single study, two chemovars, n=5, one drying temperature, one laboratory), so a dry weight concentration can rise while the absolute quantity falls. No published study traps the volatiles carried off in the drying air, so an absolute mass balance for terpenes does not exist. Every terpene percentage published for drying is therefore a relative figure, and the first question to put to one is which denominator it uses. The thermal series comes from [20] (single study, three hemp varieties, replicates not declared in the readable parts), a UC Davis study on CBD-dominant industrial hemp, inflorescences and leaves of Pipeline, Maverick and Queen-Dream-CBD harvested in 2020, with THC declared “usually lower than 0.2 % to 0.3 %” in the introduction. The conditions were ambient air, hot air at 40, 50, 60, 70 and 90 °C, infrared for 1 minute followed by hot air at 60 °C, and infrared for 2 minutes followed by hot air at 40 °C, with ambient air and freeze-drying as the two controls. The abstract reads: “As the drying temperature increased from ambient to 90 °C, drying time reduced from 1800 to 210 min, cannabidiolic-acid conversion increased from 0.2%-14.1%, and total terpene retention decreased from 82.1%-29.9%.” The 82.1% is the retention of ambient air drying and 1800 minutes is its duration, while 29.9% and 210 minutes belong to the 90 °C run. The times at the intermediate temperatures sit in Table 1, which is not accessible. The 0.2% to 14.1% is a conversion rate, the fraction of CBDA decarboxylated to CBD, and not a content by weight. Ambient air, the mildest condition in the series, already loses about a fifth of the terpenes. The floor of the study sits below 29.9%, because the conclusion reports terpene losses of up to 72.3% for the sequential infrared and hot air treatment, around 27.7% retention, so the largest loss belongs to the radiant step and the 90 °C air run comes second. On the cannabinoid side the ranking reverses. Total CBD retention lies between 83.8% and 98.6% across “drying methods and conditions” (abstract), and the low end is that same sequential infrared case, whose CBD loss the conclusion puts at “up to 16.2 %”. The graphical abstract states that “Hot air temperature did not have significant influence on the total CBD content”. Heating the air costs terpenes and converts CBDA while leaving total CBD intact, and the infrared step reaches the cannabinoids as well. The matrix carries leaf as well as flower, and the introduction puts terpenes at 0.125% to 0.278% in the leaves against 1.283% to 2.141% in the inflorescences on a dry basis, so the mixture dilutes the terpene base by close to an order of magnitude. Carrying these figures to THC-dominant medicinal flos is our inference and not a result of the paper. The authors do not say, in any readable part of the paper, what they normalised retention against. Methods and tables sit behind the paywall, while the abstract, the highlights, the full introduction, the conclusion, two section extracts and five figures are public. One constraint survives, since ambient air scores 82.1% and cannot be the reference set at 100%, which leaves the fresh material or the freeze-dried control. Until someone reads the equation, we do not claim that Chen and colleagues measured recovery against the starting mass. That is enough to settle the apparent conflict with [10], who publish absolute concentrations on dry weight and no ratio. In chemovar 240 the total terpenes rise by 25% to 30%, the individual sesquiterpenes by 20% to 50% and the sesquiterpene total by 30% to 40%, while in Gen12 the sesquiterpene total rises by 15% to 23%, and sesquiterpenes gain in relative concentration under every method tested. In Gen12 the monoterpene total holds under controlled atmosphere, nitrogen and open air, with “a relative statistically non-significant decrease of only 3-14%”, while under the 5% CO2 / 5% O2 / 90% N2 mixture the same authors measure “a notable relative increase of 26% in the total monoterpene content”. The two papers answer different questions, so the numbers do not contradict each other. Hot drying lowers recoverable terpenes and concentrates what survives on the dry residue, and anyone comparing two studies without checking which of the two quantities they measured reaches opposite conclusions from the same data. No study demonstrates sesquiterpene synthesis after harvest, and the parsimonious reading of any rise on dry weight is that monoterpenes leave first and the sesquiterpene fraction gains share, an ordering reported by three titles from three groups across drying and storage, [10], [56] and [72].

Freeze-drying, where a review claim and a metabolomic study diverge

[4] states that "The best method for drying medicinal

Cannabis is vacuum freeze-drying because it retains a maximal number of active compounds like the aroma and full flavour", in a review that reports no primary data of its own. [72] (single study, three varieties, n=3 biological replicates) measures the opposite. Of 149 metabolites, 69 differ significantly between tray drying at 18 °C and 55% RH for six days and freeze-drying at -60 °C and 0.01 mbar for 24 hours, with 45 lower in the freeze-dried material and 24 higher. Terpenes account for 28 of the 45, and three of the 45 are cannabinoids, the ones named in the text being the neutral CBG and THC. We read that pattern as suppressed decarboxylation combined with the loss of volatiles, an interpretation the authors do not state. The material went into liquid nitrogen and stayed overnight at -80 °C before sublimation, and the study does not separate the effect of that pre-freezing from the effect of the freeze-drying. The three varieties are two chemotype III CBD and one chemotype IV CBG, so no one has run this comparison on THC-dominant material. [20] is no counterweight, because freeze-drying serves as a control there and the paper claims nothing for it. [19] (single study, industrial hemp, cultivar and replicates not declared, verified from the abstract), from the same UC Davis group as [20], reports total terpene retention between 18.3% and 71.1% across its infrared and hot air conditions without saying which condition gives which value, so the interval is usable only as an interval. [30] (single study, Box-Behnken design, replication not verified, figures read from the publisher abstract) reports a 43% terpene reduction at the optimised point of a microwave and infrared pretreatment applied before drying at 25 °C and 50% RH, an optimum computed against objectives that did not include terpene retention. A terpene percentage stated without its metric, its denominator and its analytical method has no comparison value against anyone else’s, and the supply contract has to name all three.

Three groups measured fungal load before and after a drying cycle, and in each of them the load moved with the time the material spent wet. A fourth compared drying methods without a baseline and found the same ordering. Uziel and colleagues dried a THC-dominant medical chemovar, Infiniti, by solid-state microwave against a traditional reference run “under controlled temperature and humidity conditions (16 C, 50% relative humidity) for 10 days”, with total yeast and mould counts run by an external laboratory on fresh material and after each method (single study, one chemovar, one comparison condition, one research group). They report that “Traditional drying resulted in a two-log increase in the TYMC. This is a known issue, as during the long drying time, water is available to microorganisms to promote growth” [80]. After microwave drying the count “was similar to that of fresh inflorescences” [80]. Ten days at 16 °C produced two log of growth, so the setpoint alone does not determine the microbiological outcome. Baek and colleagues dried industrial hemp with hot air at 75 °C for eight hours, against ambient air at 26 °C and freeze drying (single study, two cultivars, three plants per cultivar in triplicate, one research group) [9]. Hot air leaves yeast and mould counts up to 2 log below the other two methods on the cultivar Hempress, across values of 0.23 to 3.48 log CFU/g dry mass, with about 1 log of spread on Wild Bourbon. That is not a log reduction against fresh material, since Baek and colleagues report no starting count, and the authors qualify the treatment, writing that hot air “could partly pasteurize the hemp biomass but is not considered adequate to ensure microbial safety” [9]. The two results sit at different evidentiary levels, an increase measured against a fresh baseline on one side and a spread between methods with no baseline on the other. Birenboim and colleagues held both arms at 15 °C, running controlled atmosphere below 10% relative humidity for six days against open-air at 50 to 55% for fourteen days plus one day of silica gel curing (single study, two chemovars, fungal data on chemovar 240 only, n=5 per treatment, one research group) [10]. Open-air took the load from 2.8 ± 0.8 to 31 ± 19 × 10⁶ CFU/g dry mass (p<0.0332) and Alternaria alternata from 400 ± 200 to 15,000 ± 2,800 relative units by real-time qPCR (p<0.0001), with about 80% of chemovar 240 inflorescences heavily infested by day 12, while no visible mould appeared under controlled atmosphere. Birenboim and colleagues varied humidity and duration together, so their design does not separate the two variables. Both act through the water available to the organisms. The direction of the effect holds across those four groups. The magnitude does not travel with it, because each group reports it on its own terms, as a log increase against fresh material, as a spread between drying methods with no baseline, or as counts on a different medium in different units. The temporal profile comes from one observer. Punja and colleagues write that “Following an initial increase in TYM levels during the first 3 days of drying, the populations dropped significantly to around 2,000 cfu/g on day 6” (single study, twelve sampling positions per day across six days, three repetitions, one research group), in a drying room at typically 21 to 23 °C over five to seven days [65]. They place the peak away from the end of the cycle without establishing the mechanism of the rise, so a finished product that conforms does not by itself demonstrate control of the process that made it. No regime in this review works as a validated reduction step, and drying to low water activity stops the growth of most fungi without killing Penicillium and Aspergillus, which survive between aw 0.62 and 0.70, a range Gwinn and colleagues report from general mycology and not from a measurement on cannabis (narrative review, co-authored by Punja) [42].

Suitability of the counting method

Six studies in this review publish quantitative fungal counts, namely [65], [80], [10], [9], [52] and [46]. I searched their full texts for method suitability, suitability of the counting method, recovery, neutralization, neutralizer, growth promotion, 2.6.12, 2.6.13, USP <61> and USP <62>, then repeated the search for suitability independently on the six PMCIDs in Europe PMC, with a positive control that returned hits. Of the six, none documents verification of the suitability of the counting method on the cannabis matrix. One of the six cites the chapter. Uziel and colleagues state that samples were tested “according to USP <61>/Ph. Eur 2.6.12” (Materials and Methods, “Microbiological assay”) [80]. Citing the chapter is not documenting the test it prescribes. The section “Suitability of the counting method in the presence of the product” requires a demonstration, on that matrix, that the count in the presence of the product does not depart from the control without product, once the antimicrobial activity of the product has been neutralised or removed [23]. Recovery, neutralization and growth promotion do not appear in that paper. The contract laboratory may have run the test, and the reader cannot tell from what Uziel and colleagues publish. These six studies do not document the verification, which is not the same as not having done it. The other five do not invoke the pharmacopoeia for microbiology. Punja and colleagues invoke it twice for physical parameters, moisture by USP 731 and water activity “following the procedure of USP 922” (§2.8), while their count runs by spread plating on potato dextrose agar with streptomycin [65]. Birenboim and colleagues and Baek and colleagues work on 50 mg aliquots, and Baek refers the method to the plate manufacturer, “following the manufacturer’s instructions” (§3.6) [10] [9]. Jerushalmi and colleagues inoculate Botrytis cinerea onto the inflorescence before drying, which is a challenge on the plant material and not a spike on the sample to verify recovery [46]. McKernan and colleagues come closest to the problem and state it without setting out to test it. They write that “Cannabis is a unique matrix, in that antibiotic cannabinoids can make up to 20% of the flowers' weight” (Introduction), then measure “a ten-fold difference in recovery between these media” across DRBC, PDA and PDA with chloramphenicol (Discussion), the authors being employees of the company that manufactures the qPCR reagents used in their work, an interest the paper declares in the byline and in a competing interests statement (single study, one research group) [52]. They name the antimicrobial matrix and show an order of magnitude between media, and the suitability test with a known inoculum is absent from their work too. A log change in total yeast and mould count is a difference between two counts whose recovery on this matrix nobody has documented. The direction of the effect survives, carried by convergence across independent groups. No one has demonstrated the suitability of the counting method on this matrix, so the magnitude in colony forming units stands unverified. Specified micro-organisms leave a gap of a different kind. None of the six studies performs the tests for Escherichia coli, Salmonella, bile-tolerant Gram-negative bacteria or Aspergillus as a specified organism [24]. Punja and colleagues identify A. ochraceus and A. fumigatus by taxonomic characterisation of colonies from the total count, not as an absence test [65]. These studies do not fail that requirement, because none of them addresses that test plan.

The regulated panel and the Fusarium gap

Compliance testing on cannabis covers five molecules, aflatoxins B1, B2, G1 and G2 plus ochratoxin A. On the largest published dataset of mycotoxin prevalence in legal cannabis (single study, n=328,682 flower samples and 114,579 concentrate samples from eleven US states; abstract verified, full text not consulted), Boyar and colleagues report that “overall failure rates were low (0.016% for flower and 0.017% for concentrates)”, with ochratoxin A the most frequently detected molecule in both matrices. Detection rates diverge between laboratories far more than failure rates do. “Reported detection rates across all laboratories were 3.28% for flower and 8.32% for concentrates. After excluding laboratories with disproportionately high detection rates, adjusted detection frequencies declined to approximately 0.1% for flower and 0.12% for concentrates” [14]. A factor of thirty separates those two numbers, and it follows from the authors' decision on which laboratories to exclude, so it measures analytical variability between laboratories rather than the product. The same authors ask “whether current mycotoxin panels reflect the most relevant risk drivers in cannabis products” [14]. Two surveys of seized material bear on that question. Buchicchio and colleagues analysed 142 illicit samples from Luxembourg for two analytes and found that “No AF contamination (LOD = 0.04 microg/kg) was detected in any of the samples analyzed. OTA however was detected in about one-third of the samples with an average concentration of 4.30 microg/kg (range from 1.02 to 16.21 microg/kg)” (single study, n=142, seized material of unknown post-harvest history, aflatoxins and ochratoxin A the only analytes tested) [15]. The highest value they measured stays below the 20 ppb limit many jurisdictions apply. Stephens and colleagues widened the panel on 118 dried flower samples seized in Arizona and California, and report that “Sixteen percent of 118 seized cannabis samples had detectable levels of mycotoxins and fungal metabolites”, while “Only one sample contained one of the five regulated mycotoxins”, ochratoxin A at 6.9 ppb, with aflatoxins B1, B2, G1 and G2 never detected (single study, n=118, seized material of unknown post-harvest history). Most of what they found were Fusarium toxins, fusarenon-X in ten samples at 500 to 1,700 ppb, diacetoxyscirpenol in three at 276 to 297 ppb, deepoxy-deoxynivalenol in two at 316 to 876 ppb, together with sterigmatocystin in three samples at 2.1 to 42.2 ppb. The fusarenon-X concentrations run two orders of magnitude above the 20 ppb limit that applies to the regulated molecules, and the authors state that no jurisdiction requires testing for Fusarium toxins. Stephens and colleagues found no visual clue on the contaminated samples, since “None of the samples showed visible signs of mold growth” [73]. The regulated panel caught one sample of 118, ochratoxin A at 6.9 ppb, while fusarenon-X reached 1,700 ppb outside it. Aguirre-Ortega and colleagues isolated a strain identified as Aspergillus westerdijkiae from inflorescences taken “after the drying process when the flower had a humidity between 12% and 14%”, a figure whose basis, wet or dry, the authors do not declare, and by whole-genome sequencing they found “the cluster of genes directly involved in OTA biosynthesis (otaA, otaB, otaC, otaR and otaD)” (single study, n=20, two genotypes, one site in Colombia; the authors did not quantify ochratoxin A, so the work establishes genetic potential and not production) [3]. Gwinn and colleagues assemble the water activity thresholds the sector quotes, and they take them from general mycology rather than from measurements on cannabis. “If sufficient drying is achieved, the aw (water activity) of cannabis tissues is reduced to <0.7, which prevents the growth of most fungi”, while “Penicillium and Aspergillus spp., can survive at aw of 0.62-0.7, this range is suboptimal for fungal growth”, and the same review adds that “it is improbable that mycotoxin levels would significantly increase in cannabis tissues maintained at a low aw” (review, thresholds cited second hand from the general mycological literature, not measured by the authors on cannabis) [42]. The drying target suppresses growth and leaves viable spores of the two genera that produce ochratoxin A, one of which Aguirre-Ortega and colleagues recovered from material at 12% to 14% moisture. The European monograph on cannabis flower carries no mycotoxin limit in its own text. A search of the monograph from Definition to Labelling returns no occurrence of mycotoxin, aflatoxin, microbiological criteria or water activity, and those attributes reach cannabis flower through the general monograph on herbal drugs and the general chapters [22].

Decontamination treatments

Frink and colleagues, at the California Department of Public Health, inactivated the four Aspergillus species named in Californian regulation with X-rays at 2.5 kGy, “at low (10^2 spores/g dried flower), medium (10^3 spores/g dried flower) and high (10^4 spores/g dried flower) levels of inoculation”, while at 2.0 kGy “some growth was detected in samples spiked with A. fumigatus and A. flavus” (single study, spike model on inoculated spores, n=21 for the chemical analysis) [39]. Jerushalmi and colleagues compared gamma, electron beam and cold plasma on one matrix. Gamma “reduced CFU levels by approximately 6- and 4.5-log fold, in uninoculated and artificially inoculated B. cinerea MC inflorescences, respectively”, electron beam “reduced TYM CFU levels by approximately 5-log-fold” on naturally contaminated commercial material, and the electron beam dose that halves yeast and mould counts was “calculated as 3.6 KGy” (single study, one matrix, sample number and cultivars not given in the abstract, full text behind a login wall) [46]. The same treatment loses 1.5 log between natural contamination and artificial inoculation, so spore load and location weigh more in the outcome than the choice of technology. Neither study at the top of the dose range reports sterility. Majumdar and colleagues, at 15 to 20.8 kGy, write of “the virtual sterilization of the plant material, as evidenced by the low levels of bacterial and fungal colony-forming units (CFUs) < 10 after gamma irradiation”, which anchors the phrase to a count threshold (single study, sample number and cultivars not given in the abstract) [51]. Rani and colleagues, searching for viable spores instead of counting colonies, report that irradiation “did not achieve complete sterilization. Viable spores of toxigenic fungal genera, such as Aspergillus, Penicillium, and Fusarium, persisted”, and that ELISA “confirmed aflatoxin, ochratoxin, DON, and T2 toxins in both IR and LP samples at variable concentrations” (single study; the gamma dose does not appear in the abstract and the full text was not accessible, so this result cannot be set against dose-defined work, and the sample number is not derivable) [68]. The two results sit together, because a count below 10 CFU/g is compatible with residual viable spores recovered by targeted search. Toxin already formed is a target distinct from the organism, and Rani and colleagues are the only group in this corpus who measured it after treatment, by a semi-quantitative method whose cross-reactivity on complex matrices calls for confirmation in LC-MS/MS. Across these studies the dose spans a factor of eight, from 2.5 kGy of X-rays [39] to 15 to 20.8 kGy of gamma [51], and no one has established a shared relationship between dose, efficacy and quality across them. The electron beam dose in [41] and the gamma dose in [68] do not appear in their abstracts, which blocks even a retrospective comparison. Goffman and colleagues kept a non-irradiated control in storage alongside the treated material, the only study in this corpus to do so. Over twelve weeks on two hemp cultivars, “Terpene content decreased by 8.4% immediately after irradiation, followed by further declines during storage, reaching 22.3% and 24.0% average losses in non-irradiated and EB-irradiated samples after 12 weeks, respectively”, and “EB irradiation caused a higher decrease in monoterpenes (10.8%) than in sesquiterpenes (2.5%)” (single study, two cultivars, hemp rather than high-THC flower, e-beam dose not given in the abstract) [41]. At twelve weeks the gap between treated and untreated material is 1.7 percentage points, so storage time carries most of the terpene loss that producers blame on irradiation. Hazekamp, whose work legitimised gamma treatment in European medical cannabis, found that “treatment did not cause changes in the content of THC and CBD” and that the effect was “limited to a reduction of some terpenes present in the cannabis, but keeping the terpene profile qualitatively the same” (single study, four varieties, single author, no replicates declared in the abstract). The conclusion is conditional, that “gamma irradiation of herbal cannabis remains the recommended method of decontamination, at least until other more generally accepted methods have been developed and validated” [43]. The 10 to 20% terpene loss that the sector attributes to that work, and its dose of 10 kGy, come from a later paper, where Frink and colleagues write that “Hazekamp showed that dried cannabis flowers exposed to 10 kGy of gamma irradiation did not alter THC and CBD levels or terpene composition qualitatively but terpene content was reduced by 10-20%” [39]. Nestel and colleagues place ozone and plasma-processed air outside this evidence base, describing them as methods that “are beginning to gain attention because of their preliminary results in effectively decontaminating food products”, results obtained on food rather than on cannabis (review; the authors' institute develops the non-thermal plasma technology the review presents as promising, and the numerical comparisons in the body of the article sit behind a login wall) [58].

Of the eleven cardinal papers verified in full for this review, one publishes water activity values measured on cannabis. Uziel and colleagues report aw 0.495 ± 0.002 after ten days of traditional drying at 16 °C and 50% relative humidity, and aw 0.571 ± 0.002 after 1 h 40 min of solid-state microwave drying at a 50 °C setpoint, paired in the same table with moisture of 13.10 ± 0.21% and 16.99 ± 0.42% (single study, n=3, two conditions, one group) [80]. The first value sits below 0.55 and the second inside the 0.55 to 0.65 band the industry works to. The declared moisture basis does not hold. The authors give these figures as dry basis, and 77.03 ± 0.75% dry basis for the fresh material in the same table is arithmetically impossible, since 77% dry basis is 43.5% wet basis. That inconsistency is the paper’s own. The second paper that states it measured water activity publishes no value. Punja and colleagues write in the Methods that “Water activity measurements were also made for each sample following the procedure of USP 922”, and no per-sample value follows in the Results, where water activity appears once, in the caption of Figure 12C, which reports moisture only [65]. The aw numbers that do appear in their text are of two other kinds. One is a process target they take from a CRC Press handbook chapter that is not peer reviewed and that nobody can check without buying the volume, 12 to 14% moisture “equating to a water activity of 0.65-0.7” [16]; the moisture basis behind that equivalence stays unverifiable for the same reason. The other is a conversion the authors declare on their own material, “even after 6 days of drying to a moisture content of 12-14% (translating to aw of ca. 0.62-0.7), there were still remnants of microbial activity that could be detected on plating media” (Discussion). They assert an inverse relation between water activity and colony counts without publishing a coefficient, and they do not declare the moisture basis; the method they cite, USP <731> loss on drying, implies wet basis. Baek and colleagues mention water activity three times, qualitatively, and measure none [9]; Birenboim and colleagues do not measure it [10]. Uziel and colleagues recommend holding aw at or below 0.65 on the authority of a book chapter [55] and a non-peer-reviewed online resource [45], and not of their own experiment. The 0.55 to 0.65 range comes from a three-page voluntary specification that ASTM sells for 64 US dollars, “Standard Specification for Maintaining Acceptable Water Activity (aw) Range (0.55 to 0.65) for Dry Cannabis Flower Intended for Human/Animal Use” [7]. Its public abstract gives the rationale in both directions: “aw shall be less than 0.65 to ensure against undesirable growth of microorganisms such as mold, and shall be greater than 0.55 to ensure against physical damage (breakage) in routine handling and storage”. The lower bound is a mechanical limit, and the ASTM abstract attaches no microbiological reason to it. The companion practice places aw control in a HACCP programme as a critical control point that focuses on preventing the growth and proliferation of microorganisms, and states that it is not a kill step [5]. Neither document makes an experimental derivation on cannabis publicly visible, and whatever derivation exists sits behind the purchase price. Das and colleagues, in the main peer-reviewed review of cannabis post-harvest operations, give no numerical aw target and do not cite D8197 [31]. The USP Herbal Medicines Compendium monograph for cannabis inflorescence carries 0.60 ± 0.05, numerically the same interval [79]. Secondary sources place it in different sections, Storage in one and Specific Tests in another, and hmc.usp.org returned HTTP 403 on two paths and refused browser navigation, so the defensible reading is a storage condition and not a release criterion. Ph. Eur. 3028 carries no water activity specification, and a case-insensitive search of the extracted monograph text, from Definition to Labelling, returns zero occurrences of the term [22]. The European monograph regulates total water instead, at not more than 12.0 per cent loss on drying, determined on 1.000 g of cut or powdered drug over about 100 g of molecular sieve R at 1.5 to 2.5 kPa and 40 °C for 24 h, which makes it a percentage of the mass taken and not of dry matter. ASTM and USP converge on one interval; the European pharmacopoeia does not carry the parameter. The closest European text names the number without turning it into a limit. The Committee on Herbal Medicinal Products writes in its non-binding reflection paper on microbiological aspects of herbal medicinal products that “It is generally recognised that in products with aw below 0.60 moulds and yeasts do not proliferate”, and puts about 0.70 in the same paragraph as the value below which the great majority of moulds do not grow [35]. The same paper records that dried herbal material stored under normal conditions usually sits between 0.50 and 0.60, and states that a water activity test “cannot replace a test on TAMC and TYMC”. The thresholds that the cannabis literature quotes come from general mycology. Punja’s Discussion gives a growth minimum near 0.61, which the authors attribute to Beuchat 1983, and a 0.62 to 0.70 window for xerotolerant species [65]. Gwinn and colleagues report that “most fungi grow best at aw of 0.83-0.99” and that Penicillium and Aspergillus species “can survive at aw of 0.62-0.7, this range is suboptimal for fungal growth” [42]. That sentence describes survival at those values and a growth rate below the optimum, not an absence of growth. Al Ubeed and colleagues sit at the other end of the spread, recommending storage below aw 0.30 in the same sentence in which they name 11% w/w moisture as able to activate microbial activity, while the same review gives 11% w/w as its drying target; the review reports both figures as w/w without saying whether it means wet or dry basis [4]. Their recommendation lies a factor of two away from every primary value measured on cannabis in this corpus, and neither end of the spread rests on a primary measurement on this matrix. Among the eleven papers verified for this review, none relates water activity to cannabinoid stability. Uziel measures aw and cannabinoids and does not correlate them; Milay and colleagues study stability without measuring aw [56]. No study in this corpus runs a dose-response experiment at controlled aw on this matrix. The only published sorption isotherms for hemp floral material are on flower and leaf composite powder (single study) [59], and grinding destroys the physical structure that governs moisture transport in an intact bud; the only work on whole inflorescence is a Master’s thesis under embargo until 10 March 2027 whose repository record shows no abstract [81]. The Das review and that thesis come from the same Saskatchewan group, so both entries count as one laboratory. No isotherm exists for whole flower, and no study in this corpus compares a benchtop analyser reading, taken by Uziel at 105 °C, with the compendial loss on drying at 40 °C in vacuo for 24 h, so a manufacturer cannot read a 12% from an analyser as the compendial 12.0%. Those two gaps separate the industrial criterion from the compendial one. Both gaps sit in the evidence. Neither makes a supplier non-compliant, and neither turns 0.60 ± 0.05 into a European requirement.

The main peer-reviewed review of cannabis post-harvest operations says that “Currently, no model has yet been established for the prediction of the drying endpoint and total drying time”, and that “there are no standards for which the endpoint of drying is defined, and practices are based on word of mouth” [31]. It is a narrative review, with no stated protocol and no quality appraisal of the studies it collects. The statement is usable because the primary studies behave as it describes. Six studies run on five stopping rules, and no arithmetic converts one into another. Uziel and colleagues stopped “when samples were dry to the touch and the stem holding the inflorescences was broken easily, or when the drying time was greater than 4 h”, a sensory judgement under a four hour ceiling (single study, one chemovar, n=3, n=2 for terpenoids) [80]. Birenboim and colleagues used constant weight, anchored to a water content of 10 ± 1% for the dried inflorescence, and do not say whether that figure sits on a wet or a dry basis (single study, two chemovars, n=5) [10]. Spadafora and colleagues dried for six days, which they justify as the “typical time required for the moisture loss to reach a plateau”, and publish neither a moisture content nor a water activity anywhere in the paper (single study, three low-THC varieties, three biological replicates) [72]. Yoon and colleagues obtained their thresholds after seeing the data, “determined as the range that could be best classified through sensitivity analysis”, and propose an endpoint reached when samples with a high CBDA conversion rate occupy more than 80% of the image, at nine days of hot-air drying (single study, one cultivar, 4,707 spectral points from 90 images) [83]. Cleary and Punja both ran a fixed six-day cycle. Cleary quantified variability for total THC alone, with nothing equivalent for moisture or water activity (single study, 12 lots, 8 cultivars) [21]. Punja measures residual moisture after the six-day clock has already ended the cycle. (single study, twelve room positions sampled at daily intervals, procedure repeated three times) [65]. Yoon’s own data undercut the weight criterion, since “Although the weight and RWC remained constant during the drying process, the concentrations of CBD, THC, and their precursors varied depending upon the drying conditions” (single study, one cultivar, thresholds derived after the fact) [83]. Yoon’s thresholds come from runs that had already finished, so they cannot accept or reject a batch while it is still drying. The endpoint the industry uses is the first one on that list, and no one in this corpus has measured it. Uziel’s sentence is the written form of the stem snap test, where a technician bends a stem, listens, and decides. No study in this corpus reports the agreement between two operators applying that test to the same material, or of the same operator on two different days. The regulatory text asks for the parameters rather than the judgement. GACP Rev. 1 requires that “Individual conditions must be recorded in detail (e.g. drying temperature, duration, method)”, and fixes no value [34]. That missing reproducibility study is a gap in the literature rather than a finding against a producer, and it leaves the producer as the only available source of the number. Process analytical technology does not close the loop here. No study in this corpus controls cannabis drying in feedback on a measurement of the material. Yoon’s hyperspectral campaign is the only one that runs during drying, and it observes rather than controls, because the authors classified the spectra after the run [83]. The nearest thing to an instrumental criterion comes from the same microwave prototype, which runs its closed loop on a temperature setpoint while its authors report that “a linear relationship (correlation coefficient of 0.996) was found between the moisture content and the FPC number”, a number produced by the drying system rather than measured on the inflorescence (single study, one chemovar, n=3) [80]. The closest technical precedent sits outside the plant. Avila and colleagues mounted a MEMS Fabry-Perot NIR sensor in line on a fluid-bed dryer, followed 14 batches of placebo granules from about 33-37 wt% to 1.3-2.6 wt%, reported as weight per cent with no basis stated, with an accuracy of ±13%, and identified the endpoint from multivariate statistical process control on the spectra alone (single study, placebo granules) [8]. The transfer from placebo granules in a fluidised bed to resinous inflorescences hanging in still air, which start at more than twice that water content, is our inference. The European texts do not ask for a loop either. Neither EudraLex Annex 7 nor GACP Rev. 1 contains a single occurrence of “process analytical”, “in-line”, “real-time release” or “water activity”, and we checked both official texts word by word [33] [34]. An inspector opens the monograph. Ph. Eur. 3028 sets loss on drying at not more than 12.0 per cent, determined by method 2.2.32 on 1.000 g of the cut or powdered drug, not sieved, dried over about 100 g of molecular sieve R at a pressure of 1.5 to 2.5 kPa, at 40 °C for 24 h [22]. Two of the published figures sit above that limit and a third straddles it. Uziel’s traditional control ended at 13.10 ± 0.21% and the 50 °C microwave condition at 16.99 ± 0.42%, and the authors read both on a moisture analyser at 105 °C and declare them on a dry basis, a declaration internally inconsistent with the 77.03% the same authors report for fresh material (single study, one chemovar, n=3) [80]. Punja reports 12-14% at day six without stating a basis in the text, citing USP <731> as the method [65]. The 10 ± 1% of Birenboim carries no basis at all [10]. None of these figures came from the compendial method, no study in this corpus bridges an analyser reading to a loss on drying under monograph conditions, and several of them leave their denominator undeclared. None of these batches was dried for release in Europe, so this is not a prediction that the material would fail. Conditions that the literature recommends are not compliance benchmarks, and a producer who takes a moisture target from a paper has set it against a measurement that does not translate into the one the monograph will apply.

Cleary and colleagues sampled hang-dried flower after six days and found total THC varying by 3.1 to 6.7 percentage points w/w inside the apical stratum alone (single study, n=12 batches, 8 THC-dominant cultivars, one producer, hydroponic greenhouse). Cultivar K2 ran from 20.3 to 23.4 per cent and cultivar F1 from 20.2 to 26.9 per cent, in the nine batches sampled at the top only, on 27 to 57 analytical replicates of about 1 g each. The propagated uncertainty of their method, validated to ICH Q2(R2), is ±0.2 percentage points, and 29.6 to 41.0 per cent of individual replicates fell inside the 99 per cent confidence interval of their own batch mean [21]. They fixed the 27-replicate minimum a priori by power calculation (G*Power, alpha 0.01, beta 0.90, d 0.8). Biological variability inside one batch therefore exceeds analytical uncertainty by more than an order of magnitude, and you cannot repair a sample that fails to represent the batch by tightening analytical precision. Part of that spread follows a vertical gradient, and two independent groups report it on different chemotypes. Danziger and Bernstein write that “the cannabinoid concentrations at location V were reduced considerably in all treatments by up to 40%” (full text, PMC8468686), with position V the lowest sampled on plants of about 2.5 m, and they attribute the reduction to light penetration (single study, cultivar Topaz, chemotype III, 6 plants and 25 to 30 samples per treatment, eight architecture treatments) [27]. On THC-dominant material Cleary found top and middle strata systematically more potent than the bottom by 4.7 to 6.1 percentage points in the three batches sampled at three heights, with effect sizes of 1.71 (cultivar D15) and 2.45 (cultivar K4), and a mean difference of 2.8 percentage points between plants of the same cultivar (p<0.0001) [21]. Both groups agree on the direction. They give no common magnitude, since one works on a chemotype III cultivar in the standing greenhouse plant and the other on THC-dominant flower already dried, each at a single production site. No study in this corpus asks whether drying amplifies or flattens the gradient it inherits, and the answer decides whether you keep sampling stratified by position after drying as well. The compendial tolerance is where that variability lands. Monograph 3028 of the European Pharmacopoeia (Supplement 11.5, 07/2024:3028) requires that, where the herbal drug is intended for prescription to the patient as a medicinal product, the measured contents of total THC and total CBD do not deviate from the labelled values by more than 10 per cent in either direction [22]. Take Cleary’s F1 batch, mean 23.3 per cent, and set the label at that mean. The window then runs from 21.0 to 25.6 per cent, while the authors measured individual replicates at 20.2 and 26.9 per cent, outside it in both directions. In batches K2, B17 and P8 the individual replicates stay inside. The arithmetic and the assumption about the label are ours and not the authors', and what they show is that a single sample can fall outside the label tolerance while the batch mean sits inside it. The sampling plan therefore belongs inside the control strategy for that specification. No study in this corpus states how a batch is defined. Cleary is the only one using the batch as the unit of analysis, and Punja’s more than two thousand samples collected over three years carry no declared batch structure. None of these studies addresses what constitutes a batch (a room, a cultivar, a harvest day, a dryer load), the genealogy between rooms, blending as a variance-reduction tool, or the batch number crossing the GACP to GMP boundary. In the field’s own narrative review, which declares no search protocol, the words uniformity, gradient and sampling appear in no discussion of batch or chamber variability [31]. A sampling standard does exist. ASTM D8334/D8334M-20, Standard Practice for Sampling of Cannabis/Hemp Post-Harvest Batches for Laboratory Analyses (7 pages, subcommittee D37.03), covers harvested inflorescence and excludes processed materials, pre-harvest field sampling and large untrimmed or unprocessed lots. ASTM does not publicly disclose the prescribed sample size, the number of increments or any stratification, so we cannot establish whether the practice accounts for the gradient above without buying it [6].. You write the drying specification on room air and you verify it on product, and those are two different measurements. Punja and colleagues collected samples from twelve positions in the drying room at daily intervals over six days and repeated the experiment three times (single study, n=12 positions per day, three repetitions), stating in their Methods that “Samples were collected from 12 locations within the room”. The design carries as much as the result, because the people who ran the study did not treat a drying room as representable by one sample. The same work places the critical window at the start of the cycle and not at its end, reporting that “Following an initial increase in TYM levels during the first 3 days of drying, the populations dropped significantly to around 2,000 cfu/g on day 6” [65]. No study in this corpus measures how far the interior of a bud departs from the air around it during drying, so you determine that offset for your own room at qualification instead of borrowing it from a published value. No study in this corpus publishes a thermo-hygrometric mapping of a real atmospheric drying room. The one peer-reviewed CFD study on cannabis optimises a vacuum heat pump dryer at 20.2 kPa and 31.6 °C for the most uniform temperature distribution and the lowest static enthalpy, against a target of 10 to 14 per cent moisture on a dry basis (single study, prototype, no replicate) [60]. Fluid dynamics in a chamber held at 20.2 kPa do not transfer to a room at atmospheric pressure. Three studies do declare their chamber load. Pentamwa dries 18 kg of fresh material in 180 min [60], Baek uses 200 g per replicate [9] and Kanabus 50 ± 0.2 g per trial [47]. None of them varies the load as an experimental factor, expresses it as loading density in kg per m³ of chamber or kg per m² of rack, or relates it to uniformity and drying time. Cleary did for total THC what no study in this corpus has done for water status, with stratified sampling, an a priori sample size, effect sizes and the share of replicates falling inside the batch confidence interval. Nothing equivalent exists for moisture content or for water activity, so whether one value of either represents a batch remains unknown. This is an absence in the published record, and an auditor cannot charge a supplier with it. An auditor can ask what evidence the supplier generated in house, on its own cultivar and in its own room. The record supports one requirement, and it is narrow. You demonstrate conformity on product, in multiple recorded positions, for the whole duration of the cycle.

Das and colleagues rank curing among the most significant postharvest operations and, in the same sentence, record that the research is missing: “Although curing is one of the most significant postharvest operations, it has been overlooked and is not investigated properly” [31]. That paper is a narrative review and carries no primary data of its own. One experimental study in this corpus treats curing as a measured variable, on industrial hemp (single study, n=3 plants per cultivar, two cultivars, one research group). Baek and colleagues dried Hempress and Wild Bourbon to 1.9-6.8% moisture on a wet basis, then cured the material for three weeks in the dark at about 25 °C and 45% relative humidity, in sealed glass jars or Mylar bags [9]. Moisture then went back up: “Curing led to a 3.3% to 13.6% increase in hemp moisture, while the influence of curing method was not significant” (Results). The rise is in percentage points on a wet basis, and it takes final moisture to 5.3-16.5% wet basis depending on cultivar and container; the ambient-dried material moves furthest, from 2.9% after drying to 13.6% wet basis in glass and 16.5% in Mylar. The authors determined moisture gravimetrically at 105 °C, not by the loss on drying of [22] (1.000 g of cut or ground drug, about 100 g of molecular sieve R, 1.5 to 2.5 kPa, 40 °C, 24 h), so the top of that range sits above the 12.0% compendial limit, measured by a method the monograph does not prescribe and on hemp rather than on medicinal flos. No study in this corpus runs its own moisture figure against the compendial loss on drying, so you cannot read a 12% from a moisture analyser as a compendial 12.0%. Moisture keeps moving for three weeks after drying, so the point in the process where you draw the release sample decides the result you get, and none of these figures settles compliance with the monograph limit. The chemical cost of the operation is small. Additional decarboxylation over the three weeks stayed below 2%, except in the hot-air dried samples, against 8.5-9.0% produced by eight hours of hot air at 75 °C in the same study [9]. The authors did not measure terpenes and list that as work still to be done, although the terpene effect is how they themselves define curing. The one variable they did measure, water uptake, went up, and higher water content favours microbial growth. The larger terpene loss happens after drying, in storage. Milay and colleagues followed whole and ground inflorescences of two chemovars, THC-dominant 'Lemon Kush' and CBD-dominant 'Golan', in the dark at -80, -30, 4 and 25 °C for a year, sampling at 0, 4, 8 and 12 months (single study, n=3 per condition, one research group): “the average terpenoid concentration decreased by more than 50% at t4 as opposed to an average 26% loss for all the biosynthesized phytocannabinoids at t12” (Results) [56]. Terpenes are the early indicator and cannabinoids the late one. The same asymmetry appears in an electron beam study on industrial hemp from an unrelated group, where cannabinoid contents remained stable throughout storage while the paired untreated control lost 22.3% of its terpene content over twelve weeks (single study, two cultivars, replicate number not reported) [41]. Two results contradict common handling practice. Deep freezing was the worst condition tested for aroma: “a storage temperature of -80 °C generally resulted in the greatest decline in terpenoid concentrations for both whole and ground samples” (Results) [56], a single study that no other work in this corpus has replicated. Ground samples held lower terpenoid concentrations than the matching whole samples, while THCA fell by about 33% in type I inflorescences, whole and ground alike, after a year at 25 °C, against about 85% in the DMSO and ethanol extracts at the same temperature [56]. Fairbairn and colleagues published the mechanism fifty years ago, reporting that minimising damage to the glands reduced oxidation losses because the glands act as well filled and well closed containers (single study, nine herbal cannabis samples and two resins, consulted in abstract only) [36]. That same work names the dominant variable, and no study in this corpus has requantified it. Fairbairn and colleagues report light, and not direct sunlight, as the single most important factor in cannabinoid loss, call the effect of temperature up to 20 °C insignificant, and record significant losses from oxidation in air [36]. The recent storage studies in this corpus, Milay 2020 and Spadafora 2024 among them, all run in the dark, so the dominant factor stays constant and unmeasured. Fairbairn also reports that loss of tetrahydrocannabinol under light does not raise cannabinol, while oxidation in air in the dark does, so CBN traces the oxidative route and not storage history in general [36]. Packaging choices trade one attribute against another. Over six months at 18 °C, 60% relative humidity and in the dark, a dark glass bottle retained the initial volatile profile better than an open tray or a closed high-density polyethylene box, and “the glass bottle storage condition causes formation of neutral cannabinoids at the expenses of the highly priced acid forms”, while the box and the tray held the initial concentrations of THC, CBC, THCA and CBDA (single study, n=3 biological replicates, three low-THC varieties, two chemotype III CBD-dominant and one chemotype IV CBG-dominant, one research group) [72]. The trial ran in the dark, so the glass advantage does not come from light shielding. Birenboim and colleagues packed under vacuum and measured both attributes on the same material: over 130 days at 4 °C, THCA loss ranged from 23% in the reference packed in air to 4% under vacuum combined with a pre-harvest hexanoic acid treatment, a decline the authors report as not statistically significant, while beta-myrcene fell by 37% in that same arm (single study, n=5 per arm, one chemovar, one research group) [10]. None of this is a stability study. Milay 2020 comes closest and uses laboratory containers; Spadafora 2024 runs six months on low-THC varieties, and its authors did not design it as a stability study. This corpus holds no ICH Q1A study on medicinal flos in its real primary packaging, so only the manufacturer can generate the shelf life of a batch in the pack it sells.

Four European texts bear on the drying of medicinal cannabis flower, and none of them fixes a temperature, a duration or a residual moisture target for the process. Industry summaries report requirements these texts do not contain, so a supplier’s obligation and an auditor’s preference separate as soon as you read the texts themselves.

The boundary, and the duties the two GMP texts impose

Annex 7 of EudraLex Volume 4 (Brussels, 1 September 2008, revised version operative from 1 September 2009, never revised since) places cutting and drying at the GACP/GMP boundary through a table row reading “Cutting, and drying of plants, algae, fungi...”. The asterisked note attached to that row reads, verbatim, “Manufacturers should ensure that these steps are carried out in accordance with the marketing authorisation/registration. For those initial steps that take place in the field, as justified in the marketing authorisation/registration, the standards of Good Agricultural and Collection Practice for starting materials of herbal origin (GACP) is applicable. GMP is applicable to further cutting and drying steps.” (table note) [33]. The boundary moves, and the applicant justifies its position case by case in the dossier. The text nowhere says where an initial step ends and a further step begins. Cannabis, process analytical technology, water activity, in-line and real-time measurement appear nowhere in it, and it uses “should” 27 times against “must” twice [33]. Annex 7 demands specification and monitoring. Section 8 asks that “when a dried plant is used, the drying system should be specified” and that the water content for herbal substances be “determined in accordance with the European Pharmacopoeia”. Section 9 asks that the processing instructions “include drying time and temperatures, and methods used to control cut size or particle size”. Section 4 states that storage may require special conditions of humidity, temperature or light protection, and that the manufacturer provide and monitor them [33]. The GACP guideline EMA/HMPC/246816/2005 Rev. 1, adopted by the HMPC on 9 July 2025, hardens much of this language from “should” to “must” without fixing a single numerical value. Section 12 requires that primary processing “must be carried out as soon as possible after harvesting”, and that “Attempts must be made to achieve uniform drying of the medicinal plant and thus avoid mould formation and to maintain quality”. The sentence that governs process control reads “The drying conditions such as maximum temperature, duration and air circulation must be selected taking into consideration the medicinal plant part to be dried, such as root, leaf or flower, and the nature of its active constituents, such as essential oils. Individual conditions must be recorded in detail (e.g. drying temperature, duration, method).” (section 12) [34]. The duty runs to selecting the parameters and recording them, and the guideline prescribes none of them. It never mentions cannabis, and it hands drying and cutting back to GMP in an elastic form, “In some circumstances drying and cutting should be performed according to EudraLex Volume 4 GMP part I or II (refer the GMP Table in Annex 7)”, so the cross-reference sends the reader back to the same Annex 7 table [34].

Monograph 3028 and its silences

Ph. Eur. monograph 07/2024:3028 Cannabis flower (Supplement 11.5, applicable from 1 July 2024) regulates the result. It sets loss on drying at not more than 12.0 per cent by method 2.2.32, determined on 1.000 g of the cut or powdered drug, unsieved, dried over about 100 g of molecular sieve R at a pressure between 1.5 and 2.5 kPa, at 40 °C for 24 h. It sets total cannabinol at not more than 1.0 per cent, and, where the drug is prescribed as a medicine, it requires measured total THC and total CBD within ±10 per cent of the labelled values [22]. The monograph declares neither a wet nor a dry basis for the 12.0 per cent, and it has no need to. Loss on drying is a mass loss measured under stated conditions on the sample as presented, and not a moisture content referred to a declared basis. A case-insensitive search of the whole monograph, from Definition to Labelling, for microbio, aflatox, mycotox, pesticid, water activity, 5.1.4, 5.1.8, 2.6.12 and 2.6.13 returns zero occurrences [22]. Those attributes reach cannabis flower through the general monograph on herbal drugs and through the general chapters, and not through 3028, which carries no water activity specification. You draw the comparison with the United States on that silence. The USP Herbal Medicines Compendium monograph Cannabis Species Inflorescence (Final Authorized Version 1.0) sets loss on drying at 10.0 per cent, and content at 90 to 110 per cent of the labelled amount for total THC and total CBD [79]. On label tolerance the two pharmacopoeias coincide at ±10 per cent, and the 80 to 120 per cent range still quoted in the literature comes from the superseded Proposed For Comment version, which carried the 2020 recommendation of the USP Expert Panel [79] [71]. The divergence sits in the loss on drying limit, 10.0 against 12.0 per cent. The USP method conditions are not publicly accessible, so the comparison holds between the two limits and stops there. The Proposed For Comment version described a determination under vacuum at 40 °C for 24 h, chosen “to avoid potential loss and other degradation not related to loss of water (e.g., from decarboxylation of acidic cannabinoids and loss of terpenes)” [79]. Ph. Eur. 3028 reaches the same combination of low temperature and reduced pressure without stating a reason. A loss on drying pushed harder on cannabis returns water plus volatilised terpenes plus the carbon dioxide released by decarboxylation of the acidic cannabinoids, and 40 °C under vacuum keeps the three apart. The USP monograph also carries a water activity figure, 0.60 ± 0.05, reported for storage. Whether it operates as a release criterion or only as a storage condition cannot be settled without the primary text, because the available secondary reports place it in different sections of the monograph [79]. Numerically it is the same interval as the three-page voluntary standard ASTM D8197-22, Standard Specification for Maintaining Acceptable Water Activity (aw) Range (0.55 to 0.65) for Dry Cannabis Flower Intended for Human/Animal Use, which states its rationale in both directions, aw “less than 0.65 to ensure against undesirable growth of microorganisms such as mold, and shall be greater than 0.55 to ensure against physical damage (breakage) in routine handling and storage” [7]. The lower bound protects the flower from crumbling in handling and has nothing to do with microbiology. An industry standard and a non-European pharmacopoeia converge on a criterion the European pharmacopoeia does not regulate. No European water activity threshold exists elsewhere either. A fifth text, non-binding, the HMPC reflection paper EMA/HMPC/95714/2013 adopted on 5 May 2015, records that “It is generally recognised that in products with aw below 0.60 moulds and yeasts do not proliferate”, and in the same paragraph puts at about 0.70 the value below which the great majority of moulds do not grow. Both statements are descriptive, and the same text says that the determination “cannot replace a test on TAMC and TYMC” [35]. Industry summaries quote the 0.60 and omit the 0.70, which turns a descriptive statement into a European limit the reflection paper does not contain.

Chapter 5.1.8, and the opinion of one Land

Ph. Eur. 5.1.8 (04/2019:50108) carries a title and a structure built for herbal medicinal products for oral use and for the extracts used in their preparation. Its category B, the one at issue for flower, asks TAMC 10⁴ CFU/g with a maximum acceptable count of 50,000, TYMC 10² with a maximum of 500, bile-tolerant Gram-negative bacteria 10², absence of E. coli in 1 g and absence of Salmonella in 25 g. The chapter carries an explicit flexibility clause, which allows less stringent criteria where the typical level of contamination makes those figures unattainable, on the basis of a risk assessment covering the qualitative and quantitative characterisation of the contamination and the intended use. Its cross-reference to 5.1.4 for other routes of administration sits inside the EXTRACTS paragraph, so a hostile reader can object that the sentence speaks about extracts [25]. Applying a chapter written for oral use to an inhaled flower is an interpretive extension, and the chapter itself does not argue it. One supervisory authority applies the extension anyway. The Merkblatt of the Hessisches Landesamt für Gesundheit und Pflege (Abteilung V Pharmazie, Stand 16.07.2026) starts from a supervisory finding, that steps subject to GMP are in part carried out outside GMP-certified sites, and it attributes that to divergent readings of the rules, in particular on initial drying under GACP conditions. It holds that drying under controlled conditions counts as a critical manufacturing step and its parameters as critical process parameters, and it requires category B unless the container and the certificates supplied to pharmacies explicitly exclude inhalation. On the Annex 7 exception it writes, “Bisher ist dem HLfGP kein Verfahren bekannt, dass nachweislich dazu geführt hat, dass Blüten, die tatsächlich noch als GACP-Blüten betrachtet werden können, ohne Qualitätsverluste über weite Strecken transportiert, gelagert, eingeführt und weiter prozessiert werden können.” The authority knows of no such procedure, and the exception closes by interpretation rather than by text. The document qualifies itself as an opinion, “nach Auffassung des HLfGP”. It offers no scientific evidence for the critical-step qualification [44]. It is the position of the supervisory authority of one German Land, in a country where each Land supervises manufacture on its own territory. It is not EU law and not an EMA position, and it binds no one outside Hesse.

Access note

The Ph. Eur. texts were read on complete third-party copies, because the EDQM site returns 403; monograph number, supplement and date of application of 3028 are corroborated independently, and the 5.1.8 figures come from edition 10.0 (04/2019:50108) and need a check against the current edition before use in a signed dossier. The USP HMC monograph on hmc.usp.org was not accessible, so its figures, including the 10.0 per cent limit, the water activity value and the description of the Proposed For Comment method, rest on qualified secondary industry reporting, and neither the version number nor the posting date is confirmed on the primary document. Sarma et al. 2020 was not consulted in full text. The ASTM standard was verified on the official ASTM product record, designation, title, version and scope, and not on the full three-page text. Annex 7, the GACP guideline and the HLfGP Merkblatt were read on the issuing authorities' own PDFs; secondary sources circulate a Merkblatt version dated 08.06.2026 against the 16.07.2026 of the primary file, so the current version needs checking at the moment of use.

The preceding sections describe effects, and none of them sets a limit. No binding European text fixes a numerical process parameter for drying medicinal cannabis flower. The guideline leaves the choice to the manufacturer and requires the manufacturer to record it. Section 12 of the GACP guideline, in Revision 1 adopted on 9 July 2025, requires that maximum temperature, duration and air circulation “must be selected taking into consideration the medicinal plant part to be dried” and the nature of its active constituents, and that “Individual conditions must be recorded in detail (e.g. drying temperature, duration, method)” [34]. Annex 7 says the processing instructions should include drying time and temperatures (section 9), that storage may require special conditions of humidity, temperature and light protection, which the manufacturer provides and monitors (section 4), and that where a dried plant is used the drying system should be specified, together with “the water content for herbal substances, determined in accordance with the European Pharmacopoeia” (section 8). That annex carries “should” twenty-seven times against “must” twice [33]. The obligation is to select the conditions and record them in detail. Faced with that obligation, the auditor asks a different question. A supplier cannot be non-compliant with a threshold no European text contains, and an auditor who invents one loses the argument in the first exchange. The question that survives contradiction concerns the basis of the choice, namely what evidence the manufacturer used to set this parameter, and how the manufacturer shows it holds for this cultivar, this room and this load. Table 2 sets out the matrix, one row per block of evidence, ending in the question and its declared basis. Three kinds of question appear there. The first kind has a direct normative hook and can support an observation: the recording of the individual drying conditions [34], the specified drying system and the water content determined according to the European Pharmacopoeia [33], the monitoring of storage conditions [33], and the suitability of the microbiological counting method [23]. The last is the strongest hook available. Of the six studies in this corpus that quantify fungal load, [65], [80], [10], [9], [52] and [46], none documents the verification of counting method suitability on the cannabis matrix. One of them, [80], assigns the count to an outside laboratory and cites the chapter, “according to USP <61>/Ph. Eur 2.6.12” (Materials and Methods, “Microbiological assay”), without reporting the verification that chapter prescribes. They do not document it, which is weaker than saying nobody performed it, because a contract laboratory may hold the data. The study closest to the problem raises it without proving it, writing that “Cannabis is a unique matrix, in that antibiotic cannabinoids can make up to 20% of the flowers' weight” (Introduction), while the same authors measure “a ten-fold difference in recovery between these media” across DRBC, PDA and PDA with chloramphenicol (Discussion) [52] (single group, authors employed by the qPCR reagent manufacturer). Specified micro-organisms are a separate matter. None of the six performs the tests set out in [24], so the gap there is an absent test plan and not a failed verification. The second kind has a contractual hook only. An effect is documented, no European text attaches a number to it, and the requirement belongs in the quality agreement. Transferability of a drying profile between cultivars sits here. [10] (single study, two chemovars, n=5, one institute) report that chemovar 240 lost THCA during open-air drying at 15 °C and 50 to 55 per cent relative humidity, from 9.5 ± 0.7 to 7.6 ± 0.5 DW%, p < 0.0002, while Gen12 stayed between 4.3 and 4.8 DW% with no significant change. Asking who validated the profile on this cultivar is legitimate; asking for conformity to the conditions of that paper is not. The provenance of a water activity target sits here too. The range 0.55 to 0.65 aw comes from [7], a three-page voluntary standard that costs 64 dollars and whose published rationale sets the lower bound against physical breakage in routine handling and storage rather than against microbial growth. It matches the 0.60 ± 0.05 that [79] gives for water activity during storage, while [22] carries no water activity specification and regulates loss on drying at not more than 12.0 per cent, determined on 1.000 g of the cut or ground drug at 40 °C under 1.5 to 2.5 kPa over molecular sieve R for 24 h (method 2.2.32), a loss taken against the mass as received and therefore a wet-basis figure. A producer who controls its process on water activity owes the bridge to that determination. So does the sampling point, since three weeks of curing raised moisture by 3.3 to 13.6 percentage points on a wet basis, taking final values to between 5.3 and 16.5 per cent [9] (single study, two industrial hemp cultivars, three plants per cultivar). The third kind arises from a gap in the literature and cannot generate a finding. No study in this corpus measures whether a batch that meets pesticide and heavy metal limits on fresh material still meets them on dried material once three quarters of its mass has gone (stessa correzione per le altre due occorrenze di nobody), and no published study in this corpus reports moisture or water activity resolved by position inside the same drying room. Questions of this kind open a discussion. They cannot support an observation, because the absence of a study proves neither that the problem exists nor that it does not. A gap in the science is not a supplier non-conformity.

The research agenda

The same gaps set the agenda for the work ahead. No published study in this corpus dries THC-dominant medicinal flos in a European GMP drying room. None varies temperature, relative humidity and air velocity together in a full factorial design. The closest, [49] (single study, three floral hemp cultivars, CBD-dominant, replicate number not reported), crosses three temperatures with two air velocities and leaves the relative humidity of the air undeclared. No published work maps temperature and humidity across a loaded drying room, and nobody has quantified intra-batch or intra-chamber uniformity of moisture, although total THC varies by 3.1 to 6.7 percentage points w/w within the apical layer alone against ± 0.2 points of analytical uncertainty [21] (single study, 12 batches, 8 THC-dominant cultivars, n = 27 to 57 replicates per batch, one producer). No sorption isotherm exists for whole flower, since the only published one covers composite flower and leaf powders [59], and grinding destroys the structure that governs moisture transport in an intact bud. No study bridges an in-process moisture reading to the compendial loss on drying, and none in this corpus determines moisture by the compendial method. No rate constant for decarboxylation exists at drying temperature, because the published kinetics start at 80 °C [82], while [48] (single study, one CBD-dominant cultivar, t0 against 15 days) and [9] show the reaction running at 20 to 25 °C without measuring its rate. No study links water activity to cannabinoid stability. Nobody has measured how far two operators diverge on the endpoint in industrial use, dry to the touch with the stem breaking easily, which [80] applies under a four-hour cap. No cultivar in this corpus appears in two independent laboratories, so cultivar effect and laboratory effect stay confounded, and the claim that an optimum is cultivar specific is today indistinguishable from the claim that every laboratory obtains different results. This agenda addresses the research community and any producer willing to publish. It does not address the supplier under audit.

1. [1]Addo PW, Chauvin-Bossé T, Taylor N, MacPherson S, Paris M, Lefsrud M (2023). Freeze-drying Cannabis sativa L. using real-time relative humidity monitoring and mathematical modeling for the cannabis industry. Industrial Crops and Products 199:116754. DOI: https://doi.org/10.1016/j.indcrop.2023.116754. [abstract only]
2. [2]Addo PW, Gariepy Y, Shearer M, Taylor N, MacPherson S, Raghavan V, Orsat V, Lefsrud M (2024). Microwave-assisted hot air drying of Cannabis sativa: Effect of vacuum and pre-freezing on drying kinetics and quality. Industrial Crops and Products 218:119015. DOI: https://doi.org/10.1016/j.indcrop.2024.119015. [abstract only]
3. [3]Aguirre-Ortega E, Henao-Lopez M, Vargas-Alzate JA, Castro-Rios K, Montoya-Estrada CN (2024). Identification of Aspergillus westerdijkiae and its potential risk of Ochratoxin A synthesis in Cannabis inflorescences. Scientia Agropecuaria 15(1):45-54. DOI: https://doi.org/10.17268/sci.agropecu.2024.004.↩︎
4. [4]Al Ubeed HMS, Wills RBH, Chandrapala J (2022). Post-Harvest Operations to Generate High-Quality Medicinal Cannabis Products: A Systemic Review. Molecules 27(5):1719. DOI: https://doi.org/10.3390/molecules27051719.↩︎ a↩︎ b
5. [5]ASTM International, Standard Practice for Determination of Water Activity (aw) in Cannabis Flower, ASTM D8196-22, subcommittee D37.04, 2022 (previous versions D8196-18, D8196-20). [full text not consulted]↩︎
6. [6]ASTM International, Standard Practice for Sampling of Cannabis/Hemp Post-Harvest Batches for Laboratory Analyses, ASTM D8334/D8334M-20, subcommittee D37.03, 1 December 2020. [full text not consulted]↩︎
7. [7]ASTM International, Standard Specification for Maintaining Acceptable Water Activity (aw) Range (0.55 to 0.65) for Dry Cannabis Flower Intended for Human/Animal Use, ASTM D8197-22, subcommittee D37.04, 2022 (previous versions D8197-18, D8197-21). [full text not consulted]↩︎ a↩︎ b↩︎ c↩︎ d
8. [8]Avila CR, Ferre J, de Oliveira RR, de Juan A, Sinclair WE, Mahdi FM, Hassanpour A, Hunter TN, Bourne RA, Muller FL (2020). Process Monitoring of Moisture Content and Mass Transfer Rate in a Fluidised Bed with a Low Cost Inline MEMS NIR Sensor. Pharmaceutical Research 37(5):84. DOI: https://doi.org/10.1007/s11095-020-02787-y.↩︎
9. [9]Baek Y, Grab H, Chen C (2025). Postharvest Drying and Curing Affect Cannabinoid Contents and Microbial Levels in Industrial Hemp (Cannabis sativa L.). Plants 14(3):414. DOI: https://doi.org/10.3390/plants14030414.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g↩︎ h↩︎ i↩︎ j↩︎ k↩︎ l↩︎ m↩︎ n↩︎ o↩︎ p↩︎ q
10. [10]Birenboim M, Brikenstein N, Duanis-Assaf D, Maurer D, Chalupowicz D, Kenigsbuch D, Shimshoni JA (2024). In Pursuit of Optimal Quality: Cultivar-Specific Drying Approaches for Medicinal Cannabis. Plants 13(7):1049. DOI: https://doi.org/10.3390/plants13071049.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g↩︎ h↩︎ i↩︎ j↩︎ k↩︎ l↩︎ m↩︎ n↩︎ o↩︎ p
11. [11]Birenboim M, Brikenstein N, Kenigsbuch D, Shimshoni JA (2025). Aquaphotomics study of fresh cannabis inflorescence: near infrared spectral analysis of water matrix structures. Analytical and Bioanalytical Chemistry 417(4):747-760. DOI: https://doi.org/10.1007/s00216-024-05685-z.
12. [12]Birenboim M, Brikenstein N, Kenigsbuch D, Shimshoni JA (2025). Comparative chemometric modeling of fresh and dry cannabis inflorescences using FT-NIR spectroscopy: Quantification and classification insights. Phytochemical Analysis 36(3):537-555. DOI: https://doi.org/10.1002/pca.3449. [abstract only]
13. [13]Birenboim M, Chalupowicz D, Kenigsbuch D, Shimshoni JA (2024). Improved Long-Term Preservation of Cannabis Inflorescence by Utilizing Integrated Pre-Harvest Hexanoic Acid Treatment and Optimal Post-Harvest Storage Conditions. Plants 13(7):992. DOI: https://doi.org/10.3390/plants13070992.
14. [14]Boyar K, Esposito M, Sukhikh P, Kahn Y (2026). Mycotoxin Trends of Cannabis Compliance Testing in Cannabis Flower and Concentrates in the United States. Clinical Therapeutics 48(9):876-882. DOI: https://doi.org/10.1016/j.clinthera.2026.04.003. [abstract only]↩︎ a↩︎ b
15. [15]Buchicchio L, Asselborn L, Schneider S, van Nieuwenhuyse A, Moris G, Schummer C (2022). Investigation of aflatoxin and ochratoxin A contamination of seized cannabis and cannabis resin samples. Mycotoxin Research 38(1):71-78. DOI: https://doi.org/10.1007/s12550-022-00449-z. [abstract only]↩︎
16. [16]Caplan D, Matzneller P, Gutierrez D (2022). Harvest and post-harvest. In: Zheng Y (ed), Handbook of Cannabis Production in Controlled Environments. New York: CRC Press, pp. 291-311. DOI: https://doi.org/10.1201/9781003150442-10. [handbook chapter, not peer reviewed; not consulted directly for this review]↩︎ a↩︎ b
17. [17]Chasiotis V, Tsakirakis A, Termentzi A, Machera K, Filios A (2022). Drying and quality characteristics of Cannabis sativa L. inflorescences under constant and time-varying convective drying temperature schemes. Thermal Science and Engineering Progress 28:101076. DOI: https://doi.org/10.1016/j.tsep.2021.101076. [abstract only]
18. [18]Chasiotis VK, Tzempelikos DA, Filios AE (2021). Assessment of constant and time-varying temperature schemes on the convective drying characteristics of hemp leaves. Case Studies in Thermal Engineering 26:101098. DOI: https://doi.org/10.1016/j.csite.2021.101098. [abstract only]
19. [19]Chen C, Wang K, Wongso I, Ning Z, Khir R, Putnam D, Donis-González IR, Pan Z (2023). Postharvest blanching and drying of industrial hemp (Cannabis sativa L.) with infrared and hot air heating for enhanced processing efficiency and microbial inactivation. Drying Technology 41(9):1454-1468. DOI: https://doi.org/10.1080/07373937.2022.2159973. [abstract only]↩︎
20. [20]Chen C, Wongso I, Putnam D, Khir R, Pan Z (2021). Effect of hot air and infrared drying on the retention of cannabidiol and terpenes in industrial hemp (Cannabis sativa L.). Industrial Crops and Products 172:114051. DOI: https://doi.org/10.1016/j.indcrop.2021.114051. [abstract only]↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g
21. [21]Cleary B, Maloney K, Toor A, Vandermeirsch G (2025). Variability of total THC in greenhouse cultivated dried Cannabis. Scientific Reports 15:25285. DOI: https://doi.org/10.1038/s41598-025-06962-2.↩︎ a↩︎ b↩︎ c↩︎ d
22. [22]Council of Europe (EDQM), Cannabis flower (Cannabis flos), European Pharmacopoeia monograph 07/2024:3028, Supplement 11.5, January 2024, applicable from 1 July 2024.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g↩︎ h↩︎ i↩︎ j↩︎ k↩︎ l
23. [23]Council of Europe (EDQM), Microbiological examination of non-sterile products: microbial enumeration tests, European Pharmacopoeia general chapter 2.6.12.↩︎ a↩︎ b
24. [24]Council of Europe (EDQM), Microbiological examination of non-sterile products: test for specified micro-organisms, European Pharmacopoeia general chapter 2.6.13.↩︎ a↩︎ b
25. [25]Council of Europe (EDQM), Microbiological quality of herbal medicinal products for oral use and extracts used in their preparation, European Pharmacopoeia general chapter 04/2019:50108, Ph. Eur. 10.0.↩︎
26. [26]Council of Europe (EDQM), Microbiological quality of non-sterile pharmaceutical preparations and substances for pharmaceutical use, European Pharmacopoeia general chapter 5.1.4. Verified on edition 7.0 (01/2011:50104), which precedes the PDG harmonisation of June 2013; the edition in force is behind a paywall.↩︎
27. [27]Danziger N, Bernstein N (2021). Shape Matters: Plant Architecture Affects Chemical Uniformity in Large-Size Medical Cannabis Plants. Plants 10(9):1834. DOI: https://doi.org/10.3390/plants10091834.↩︎
28. [28]Das PC, Baik O-D, Tabil LG (2024). Microwave-infrared drying of cannabis (Cannabis sativa L.): Effect on drying characteristics, energy consumption and quality. Industrial Crops and Products 211:118215. DOI: https://doi.org/10.1016/j.indcrop.2024.118215. [abstract only]↩︎ a↩︎ b
29. [29]Das PC, Baik O-D, Tabil LG (2025). Multi-objective optimization and modeling of microwave-infrared pretreatment on drying and quality characteristics of cannabis (Cannabis sativa L.) using response surface methodology and artificial neural network. Drying Technology 43(6):961-979. DOI: https://doi.org/10.1080/07373937.2025.2478396. [abstract only]
30. [30]Das PC, Heydari MM, Baik O-D, Zhang L, Tabil LG (2025). Enhancing drying efficiency and terpene retention of cannabis using cold plasma pretreatment. Industrial Crops and Products 226:120669. DOI: https://doi.org/10.1016/j.indcrop.2025.120669. [abstract only]↩︎
31. [31]Das PC, Vista AR, Tabil LG, Baik O-D (2022). Postharvest Operations of Cannabis and Their Effect on Cannabinoid Content: A Review. Bioengineering 9(8):364. DOI: https://doi.org/10.3390/bioengineering9080364.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e
32. [32]de Melo MMR, Aguas F, Nascimento J (2023). Tray drying of medical cannabis inflorescences at industrial scale: Kinetics measurement and modeling towards the implementation of statistical process control. Industrial Crops and Products 201:116968. DOI: https://doi.org/10.1016/j.indcrop.2023.116968. [solo metadati: nessun abstract depositato, contenuto mai letto]↩︎ a↩︎ b↩︎ c
33. [33]European Commission, EudraLex Volume 4, EU Guidelines to Good Manufacturing Practice, Annex 7: Manufacture of Herbal Medicinal Products, Brussels, 1 September 2008 (revised version operational from 1 September 2009).↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g
34. [34]European Medicines Agency, Committee on Herbal Medicinal Products (HMPC), Guideline on Good Agricultural and Collection Practice (GACP) for starting materials of herbal origin, EMA/HMPC/246816/2005 Rev. 1, adopted 9 July 2025.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f
35. [35]European Medicines Agency, Committee on Herbal Medicinal Products (HMPC), Reflection paper on microbiological aspects of herbal medicinal products, EMA/HMPC/95714/2013, adopted 5 May 2015. [non-binding]↩︎ a↩︎ b
36. [36]Fairbairn JW, Liebmann JA, Rowan MG (1976). The stability of cannabis and its preparations on storage. Journal of Pharmacy and Pharmacology 28(1):1-7. DOI: https://doi.org/10.1111/j.2042-7158.1976.tb04014.x. [abstract only]↩︎ a↩︎ b↩︎ c
37. [37]Fettoukh N, Fadil M, Stambouli H, EL Bouri A, Bouyoun T, Annemer S, Boukhaled A, Farah A (2025). Chemometric and predictive modeling of long term cannabinoid transformation in stored Cannabis sativa resin. Scientific Reports 15:33827. DOI: https://doi.org/10.1038/s41598-025-03888-7.
38. [38]Filer CN (2022). Acidic Cannabinoid Decarboxylation. Cannabis and Cannabinoid Research 7(3):262-273. DOI: https://doi.org/10.1089/can.2021.0072.↩︎
39. [39]Frink S, Marjanovic O, Tran P, Wang Y, Guo W, Encarnacion N, Alcantara D, Moezzi B, Vrdoljak G (2022). Use of X-ray irradiation for inactivation of Aspergillus in cannabis flower. PLOS ONE 17(11):e0277649. DOI: https://doi.org/10.1371/journal.pone.0277649.↩︎ a↩︎ b↩︎ c
40. [40]Gloerfelt-Tarp F, Hewavitharana AK, Mieog J, Palmer WM, Fraser F, Ansari O, Kretzschmar T (2023). Using a global diversity panel of Cannabis sativa L. to develop a near InfraRed-based chemometric application for cannabinoid quantification. Scientific Reports 13:2253. DOI: https://doi.org/10.1038/s41598-023-29148-0.
41. [41]Goffman FD, Carrera DA, Latino DARS, Cronje C, Katsir L (2025). Impact of Electron Beam Treatment and Storage Duration on Microbial Stability and Phytochemical Integrity in Hemp Flowers. Molecules 30(17):3601. DOI: https://doi.org/10.3390/molecules30173601. [abstract only]↩︎ a↩︎ b↩︎ c
42. [42]Gwinn KD, Leung MCK, Stephens AB, Punja ZK (2023). Fungal and mycotoxin contaminants in cannabis and hemp flowers: implications for consumer health and directions for further research. Frontiers in Microbiology 14:1278189. DOI: https://doi.org/10.3389/fmicb.2023.1278189.↩︎ a↩︎ b↩︎ c
43. [43]Hazekamp A (2016). Evaluating the Effects of Gamma-Irradiation for Decontamination of Medicinal Cannabis. Frontiers in Pharmacology 7:108. DOI: https://doi.org/10.3389/fphar.2016.00108.↩︎
44. [44]Hessisches Landesamt für Gesundheit und Pflege (HLfGP), Abteilung V - Pharmazie, Merkblatt zur Einfuhr, Herstellung und zum Vertrieb von Cannabisblüten, Stand 16.07.2026.↩︎
45. [45]Holmes M, Vyas JM, Steinbach W, et al. (2015). Microbiological safety testing of cannabis. Online resource, not peer reviewed. Cited here only as a secondary source relied upon by Uziel and colleagues.↩︎
46. [46]Jerushalmi S, Maymon M, Dombrovsky A, Freeman S (2020). Effects of cold plasma, gamma and e-beam irradiations on reduction of fungal colony forming unit levels in medical cannabis inflorescences. Journal of Cannabis Research 2(1):12. DOI: https://doi.org/10.1186/s42238-020-00020-6.↩︎ a↩︎ b↩︎ c↩︎ d
47. [47]Kanabus J, Bryła M, Roszko M (2024). Effect of Selected Drying Methods on the Cannabinoid Profile of Cannabis sativa L. var. sativa Inflorescences and Leaves. Polish Journal of Food and Nutrition Sciences 74(4):408-418. DOI: https://doi.org/10.31883/pjfns/195594. [abstract only]↩︎ a↩︎ b
48. [48]Kim E-S, Park S-H, Kinney CA, Olejar KJ, Corredor-Perilla IC (2024). Comparison of decarboxylation rates of acidic cannabinoids between secretory cavity contents and air-dried inflorescence extracts in Cannabis sativa cv. 'Cherry Wine'. Scientific Reports 14:16411. DOI: https://doi.org/10.1038/s41598-024-66420-3.↩︎ a↩︎ b↩︎ c
49. [49]Lumu S, Sarker NC, Hatterman-Valenti H, Hammed A, Chen B, Monono E (2025). Testing Drying Kinetic Models and the Effect of Drying on Cannabinoids of Floral Hemp (Cannabis sativa L.) in a Pilot-Scale Cabinet Dryer. Journal of Biosystems Engineering 50(1):117-129. DOI: https://doi.org/10.1007/s42853-025-00255-2. [abstract only]↩︎
50. [50]Mahmoud M, BenRejeb I, Punja ZK, Buirs L, Jabaji S (2023). Understanding bud rot development, caused by Botrytis cinerea, on cannabis (Cannabis sativa L.) plants grown under greenhouse conditions. Botany 101(7):200-231. DOI: https://doi.org/10.1139/cjb-2022-0139. [abstract only]
51. [51]Majumdar CG, ElSohly MA, Ibrahim EA, Elhendawy MA, Stanford D, Chandra S, Wanas AS, Radwan MM (2023). Effect of Gamma Irradiation on Cannabinoid, Terpene, and Moisture Content of Cannabis Biomass. Molecules 28(23):7710. DOI: https://doi.org/10.3390/molecules28237710. [abstract only]↩︎ a↩︎ b
52. [52]McKernan K, Helbert Y, Kane L, Houde N, Zhang L, McLaughlin S (2021). Whole genome sequencing of colonies derived from cannabis flowers and the impact of media selection on benchmarking total yeast and mold detection tools. F1000Research 10:624. DOI: https://doi.org/10.12688/f1000research.53467.2.↩︎ a↩︎ b↩︎ c↩︎ d
53. [53]McKernan K, Spangler J, Helbert Y, Lynch RC, Devitt-Lee A, Zhang L, Orphe W, Warner J, Foss T, Hudalla CJ, Silva M, Smith DR (2016). Metagenomic analysis of medicinal Cannabis samples; pathogenic bacteria, toxigenic fungi, and beneficial microbes grow in culture-based yeast and mold tests. F1000Research 5:2471. DOI: https://doi.org/10.12688/f1000research.9662.1. [abstract only]
54. [54]McKernan K, Spangler J, Zhang L, Tadigotla V, Helbert Y, Foss T, Smith D (2015). Cannabis microbiome sequencing reveals several mycotoxic fungi native to dispensary grade Cannabis flowers. F1000Research 4:1422. DOI: https://doi.org/10.12688/f1000research.7507.2. [abstract only]
55. [55]McPartland JM, McKernan KJ (2017). Contaminants of concern in cannabis: microbes, heavy metals and pesticides. In: Chandra S, Lata H, ElSohly MA (eds), Cannabis sativa L. Botany and Biotechnology. Cham: Springer. Cited here only as a secondary source relied upon by Uziel and colleagues.↩︎
56. [56]Milay L, Berman P, Shapira A, Guberman O, Meiri D (2020). Metabolic Profiling of Cannabis Secondary Metabolites for Evaluation of Optimal Postharvest Storage Conditions. Frontiers in Plant Science 11:583605. DOI: https://doi.org/10.3389/fpls.2020.583605.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e
57. [57]Moreno T, Dyer P, Tallon S (2020). Cannabinoid Decarboxylation: A Comparative Kinetic Study. Industrial & Engineering Chemistry Research 59(46):20307-20315. DOI: https://doi.org/10.1021/acs.iecr.0c03791. [abstract only]↩︎
58. [58]Nestel S, Hedtfeld C, Spilker U, Ehlbeck J, Guenther S, Schnabel U (2026). Sanitation of medical cannabis flowers (Cannabis sativa L., flos): comparing current technologies and non-thermal plasma novelties from safety and quality perspectives. Journal of Cannabis Research 8(1):50. DOI: https://doi.org/10.1186/s42238-026-00417-9. [abstract only]↩︎
59. [59]Oduola AA, Luthra K, Atungulu GG (2022). Determination of moisture sorption isotherms of industrial hemp (Cannabis sativa L.) flower and leaf composite powders. Industrial Crops and Products 186:115201. DOI: https://doi.org/10.1016/j.indcrop.2022.115201. [abstract only]↩︎ a↩︎ b↩︎ c
60. [60]Pentamwa C, Artnaseaw A, Nabudda K (2024). Optimal condition design of cannabis drying for vacuum heat pump dryer using computational fluid dynamic method interaction with heat transfer. Case Studies in Thermal Engineering 61:104965. DOI: https://doi.org/10.1016/j.csite.2024.104965. [abstract only]↩︎ a↩︎ b
61. [61]Perrotin-Brunel H, Buijs W, van Spronsen J, van Roosmalen MJE, Peters CJ, Verpoorte R, Witkamp G-J (2011). Decarboxylation of delta-9-tetrahydrocannabinol: Kinetics and molecular modeling. Journal of Molecular Structure 987(1-3):67-73. DOI: https://doi.org/10.1016/j.molstruc.2010.11.061. [abstract only]↩︎
62. [62]Punja ZK (2021). The diverse mycoflora present on dried cannabis (Cannabis sativa L., marijuana) inflorescences in commercial production. Canadian Journal of Plant Pathology 43(1):88-100. DOI: https://doi.org/10.1080/07060661.2020.1758959. [abstract only]
63. [63]Punja ZK, Collyer D, Scott C, Lung S, Holmes J, Sutton D (2019). Pathogens and Molds Affecting Production and Quality of Cannabis sativa L. Frontiers in Plant Science 10:1120. DOI: https://doi.org/10.3389/fpls.2019.01120.
64. [64]Punja ZK, Ni L (2021). The bud rot pathogens infecting cannabis (Cannabis sativa L., marijuana) inflorescences: symptomology, species identification, pathogenicity and biological control. Canadian Journal of Plant Pathology 43(6):827-854. DOI: https://doi.org/10.1080/07060661.2021.1936650. [abstract only]
65. [65]Punja ZK, Ni L, Lung S, Buirs L (2023). Total yeast and mold levels in high THC-containing cannabis (Cannabis sativa L.) inflorescences are influenced by genotype, environment, and pre- and post-harvest handling practices. Frontiers in Microbiology 14:1192035. DOI: https://doi.org/10.3389/fmicb.2023.1192035.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g↩︎ h↩︎ i↩︎ j↩︎ k↩︎ l↩︎ m↩︎ n
66. [66]Punja ZK, Scott C (2023). Organically grown cannabis (Cannabis sativa L.) plants contain a diverse range of culturable epiphytic and endophytic fungi in inflorescences and stem tissues. Botany 101(7):255-269. DOI: https://doi.org/10.1139/cjb-2022-0116. [abstract only]
67. [67]Rafiq H, Hartung J, Schober T, Vogt MM, Carrera DA, Ruckle M, Graeff-Hönninger S (2024). Non-Destructive Near-Infrared Technology for Efficient Cannabinoid Analysis in Cannabis Inflorescences. Plants 13(6):833. DOI: https://doi.org/10.3390/plants13060833.
68. [68]Rani M, Kaddoura MJ, Samsatly J, Chamberland G, Jabaji S, George S (2025). Detection of Mycotoxigenic Fungi and Residual Mycotoxins in Cannabis Buds Following Gamma Irradiation. Toxins 17(11):528. DOI: https://doi.org/10.3390/toxins17110528. [abstract only]↩︎ a↩︎ b
69. [69]Ross SA, ElSohly MA (1996). The Volatile Oil Composition of Fresh and Air-Dried Buds of Cannabis sativa. Journal of Natural Products 59(1):49-51. DOI: https://doi.org/10.1021/np960004a. [abstract only]↩︎
70. [70]San Nicolas M, Villate A, Alvarez-Mora I, Olivares M, Aizpurua-Olaizola O, Usobiaga A, Amigo JM (2024). NIR-hyperspectral imaging and machine learning for non-invasive chemotype classification in Cannabis sativa L. Computers and Electronics in Agriculture 217:108551. DOI: https://doi.org/10.1016/j.compag.2023.108551. [abstract only]
71. [71]Sarma ND, Waye A, ElSohly MA, Brown PN, Elzinga S, Johnson HE, Marles RJ, Melanson JE, Russo E, Deyton L, Hudalla C, Vrdoljak GA, Wurzer JH, Khan IA, Kim N-C, Giancaspro GI (2020). Cannabis Inflorescence for Medical Purposes: USP Considerations for Quality Attributes. Journal of Natural Products 83(4):1334-1351. DOI: https://doi.org/10.1021/acs.jnatprod.9b01200. [abstract only]↩︎ a↩︎ b↩︎ c
72. [72]Spadafora ND, Felletti S, Chenet T, Sirangelo TM, Cescon M, Catani M, De Luca C, Stevanin C, Cavazzini A, Pasti L (2024). The influence of drying and storage conditions on the volatilome and cannabinoid content of Cannabis sativa L. inflorescences. Analytical and Bioanalytical Chemistry 416(16):3797-3809. DOI: https://doi.org/10.1007/s00216-024-05321-w.↩︎ a↩︎ b↩︎ c↩︎ d
73. [73]Stephens AB, Rivera AB, Cahill TM, Leung MCK (2025). Evaluation of Fusarium Mycotoxins and Fungal Metabolites in Seized Cannabis in Arizona and California, 2023-2024. Environmental Health Perspectives 133(3-4):047701. DOI: https://doi.org/10.1289/EHP16028.↩︎
74. [74]Su K, Maghirang E, Tan JW, Yoon JY, Armstrong P, Kachroo P, Hildebrand D (2022). NIR spectroscopy for rapid measurement of moisture and cannabinoid contents of industrial hemp (Cannabis sativa). Industrial Crops and Products 184:115007. DOI: https://doi.org/10.1016/j.indcrop.2022.115007. [abstract only]↩︎ a↩︎ b
75. [75]Tegtmeier M, Knierim L, Schmidt A, Strube J (2023). Green Manufacturing for Herbal Remedies with Advanced Pharmaceutical Technology. Pharmaceutics 15(1):188. DOI: https://doi.org/10.3390/pharmaceutics15010188.
76. [76]Thompson GR 3rd, Tuscano JM, Dennis M, Singapuri A, Libertini S, Gaudino R, Torres A, Delisle JM, Gillece JD, Schupp JM, Engelthaler DM (2017). A microbiome assessment of medical marijuana. Clinical Microbiology and Infection 23(4):269-270. DOI: https://doi.org/10.1016/j.cmi.2016.12.001. [solo metadati: contenuto non verificato in fonte primaria]
77. [77]Tran J, Vassiliadis S, Elkins AC, Cogan NOI, Rochfort SJ (2023). Developing Prediction Models Using Near-Infrared Spectroscopy to Quantify Cannabinoid Content in Cannabis sativa. Sensors 23(5):2607. DOI: https://doi.org/10.3390/s23052607.
78. [78]Tran J, Vassiliadis S, Elkins AC, Cogan NOI, Rochfort SJ (2024). Rapid In Situ Near-Infrared Assessment of Tetrahydrocannabinolic Acid in Cannabis Inflorescences before Harvest Using Machine Learning. Sensors 24(16):5081. DOI: https://doi.org/10.3390/s24165081.
79. [79]United States Pharmacopeia, Herbal Medicines Compendium, Cannabis Species Inflorescence, Final Authorized Version 1.0, 16 June 2025. [text not consulted: figures from secondary sources]↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g↩︎ h
80. [80]Uziel A, Milay L, Procaccia S, Cohen R, Burstein A, Sulimani L, Shreiber-Livne I, Lewitus D, Meiri D (2024). Solid-State Microwave Drying for Medical Cannabis Inflorescences: A Rapid and Controlled Alternative to Traditional Drying. Cannabis and Cannabinoid Research 9(1):397-408. DOI: https://doi.org/10.1089/can.2022.0051.↩︎ a↩︎ b↩︎ c↩︎ d↩︎ e↩︎ f↩︎ g↩︎ h↩︎ i↩︎ j↩︎ k↩︎ l↩︎ m↩︎ n
81. [81]Vista ARJ (2025). Determination of Water Vapor Sorption Isotherms of Cannabis Inflorescence. Master of Science thesis, Department of Chemical and Biological Engineering, University of Saskatchewan (supervisor: Tabil L). Handle: https://hdl.handle.net/10388/16676. [text not consulted: under embargo until 10 March 2027, no abstract exposed]↩︎ a↩︎ b
82. [82]Wang M, Wang Y-H, Avula B, Radwan MM, Wanas AS, van Antwerp J, Parcher JF, ElSohly MA, Khan IA (2016). Decarboxylation Study of Acidic Cannabinoids: A Novel Approach Using Ultra-High-Performance Supercritical Fluid Chromatography/Photodiode Array-Mass Spectrometry. Cannabis and Cannabinoid Research 1(1):262-271. DOI: https://doi.org/10.1089/can.2016.0020.↩︎ a↩︎ b↩︎ c
83. [83]Yoon HI, Lee SH, Ryu D, Choi H, Park SH, Jung JH, Kim H-Y, Yang J-S (2024). Non-destructive assessment of cannabis quality during drying process using hyperspectral imaging and machine learning. Frontiers in Plant Science 15:1365298. DOI: https://doi.org/10.3389/fpls.2024.1365298.↩︎ a↩︎ b↩︎ c↩︎ d