What analytes do state cannabis programs require?
No two state programs test for exactly the same analyte set, but a pattern emerges across programs. The American Herbal Pharmacopoeia (AHP) guidelines, widely referenced by state regulators, recommend the following framework for cannabis flower:
- Total Yeast and Mold (TYM): <10,000 CFU/g
- Total Aerobic Count (TAC): <100,000 CFU/g
- Total Coliforms / Enterobacteriaceae: <1,000 CFU/g
- Bile-Tolerant Gram-Negative Bacteria (BTGN): <1,000 CFU/g
- Salmonella: absent per 1g
- Pathogenic E. coli: absent per 1g
Most state programs adopt some version of this framework. The variation comes in which analytes are included, whether they are tested as presence/absence or counted, the specific limits for counted analytes, and which product categories are subject to which tests.
California represents the strictest end of the spectrum: for inhalable products, the program requires absence of six specific targets: four Aspergillus species (A. flavus, fumigatus, niger, terreus), Salmonella, and STEC, all tested as presence/absence with any detection being a failure. Non-inhalable products require only Salmonella and STEC. There are no CFU/g limits for TYM or TAC in California's program for cannabis. It is entirely a targeted-pathogen presence/absence model.
Massachusetts represents a different approach: numeric CFU/g thresholds across TYM, TAC, coliforms, and BTGN, with Salmonella and STEC as presence/absence. Both models are defensible; they are measuring different things.
How do qPCR and culture-based methods differ, and why does it matter?
The testing methodology used by your state's licensed labs determines what a result actually means. This is one of the most underappreciated variables in cannabis compliance.
Culture-based testing grows organisms on selective agar media and counts visible colonies, reporting as CFU/g. The method's accuracy depends on which growth medium is used. Research published in PMC (2021) confirmed that DRBC medium undercounted yeast and mold compared to PDA medium, while PDA without chloramphenicol overcounted due to bacterial growth. The same sample on different media can produce counts differing by several log units. Culture also fails to detect organisms that are viable but non-culturable, and it cannot detect Aspergillus reliably; Aspergillus grows slowly, colonies clump into heterogeneous macrocolonies, and there are no commercially validated culture-based tests for Aspergillus in cannabis.
qPCR testing detects the DNA of target organisms directly, without requiring growth on media. It is faster (same-day results vs. 3 to 7 days for culture), more sensitive to stressed or slow-growing organisms, and the method of choice for targeted pathogen detection, particularly Aspergillus, STEC, and Salmonella. The limitation is that qPCR detects DNA from dead cells as well as live ones, meaning a sample that was previously contaminated and remediated may still show a qPCR signal even if no viable organisms are present. Labs that use qPCR for targeted pathogens and culture for TYM counts are using a hybrid approach that reflects each method's strengths.
The practical implication for operators: understand which method your state's licensed labs use for each analyte. A borderline result has different meaning depending on the method, and a remediation strategy that works for culture-based counts may not produce the same result on qPCR.
What is the difference between a presence/absence test and a CFU/g limit?
This distinction determines whether your product has any margin for error on a given analyte.
Presence/absence means any detection is a failure. There is no acceptable level. California's model for Salmonella, STEC, and Aspergillus is the clearest example: a single detected organism in a 1g sample fails the batch. This standard reflects the view that for inhalable products and serious human pathogens, there is no dose at which exposure is acceptable.
CFU/g limits mean the product can contain organisms up to the stated threshold. TYM at 10,000 CFU/g means that a sample with 9,500 CFU/g passes and one with 10,500 CFU/g fails. This approach acknowledges that some microbial load on dried plant material is inevitable and sets a risk-based threshold rather than demanding total absence.
The product type drives which standard applies. Inhalable products, where combustion does not reliably eliminate pathogens, receive stricter treatment in most programs than edibles or topicals. Products intended for immunocompromised patients (medical programs) often have stricter limits than recreational products.
What does sampling methodology affect about test results?
The result a lab reports reflects the specific sample it received, not the entire batch. Sampling protocol determines how representative that sample is.
Most state programs specify a sampling plan: how many samples to pull, from which parts of the batch, and at what stage. The Cannabis Safety Institute notes that trimming represents the highest-contact processing stage and the most significant contamination opportunity; samples pulled before vs. after trim can produce meaningfully different results from the same harvest batch.
Heterogeneous contamination is the specific challenge: a batch of flower may have a pocket of Botrytis-affected material that never lands in the sample that gets tested. A passing COA on a heterogeneous batch provides statistical confidence about the tested portion, not an absolute guarantee about every unit in the batch. This is why environmental controls and sanitation programs matter independently of whether a specific batch passes testing.
How does the choice of testing laboratory affect outcomes?
Multi-state compliance testing research has documented significant inter-laboratory variability in cannabis microbial testing. Labs using different media, different incubation parameters, or different qPCR assays report different results on the same sample. This variability is not fraud; it reflects the genuine methodological differences between platforms.
Research from ScienceDirect (2025) on multi-state compliance data found that high mycotoxin detection rates in some states were attributable to specific laboratory methods rather than true product contamination, with detection rates dropping substantially after excluding outlier laboratories. The same dynamic applies to TYM: a lab running Petrifilm may report a different count than one running PDA, and both results are technically accurate given their methodology.
The practical implication: if your product tests marginally near a limit at one lab, consider confirming at a second lab before making operational decisions based on a single result.