# Which Genes Actually Change How You Respond to Cannabis, and Which Ones Are Being Oversold > **Author:** Benjamin Caplan, MD (ABFM Board Certified, MA Lic. #240562, NPI: 1962722637) > **Publisher:** CED Clinic (https://cedclinic.com) > **Canonical URL:** https://cedclinic.com/cannabis-pharmacogenetics-what-genes-actually-predict/ > **Published:** 2026-09-04T22:50:20-04:00 | **Last Modified / Clinically Audited:** 2026-09-28T10:40:26-04:00 > **Clinical Summary:** Cannabis pharmacogenetics, graded honestly: CYP2C9 changes THC levels, CYP2C19 shapes CBD metabolism, and consumer response testing runs well past the data. --- CED Clinical Relevance #68Strong Clinical RelevancePatients arrive with consumer genetic reports telling them which cannabis products suit their DNA. Knowing which of those findings rest on real pharmacokinetic data and which rest on a single unreplicated association is a routine part of the consultation now. Some of the variability in how people respond to cannabis is genetic, and a small part of that is well characterized. The gap between what has been demonstrated and what is being sold as a personalized cannabis genetic report is wide enough to be worth mapping carefully. Clinical Insight | CED Clinic The best-supported finding in this field concerns CYP2C9. In 43 healthy volunteers given oral THC, people homozygous for the reduced-function CYP2C9*3 variant had a median THC area under the curve about three times higher than people with two normal-function copies, and 70 percent lower exposure to the inactive metabolite. Those carriers also trended toward more sedation. Almost everything else marketed as cannabis genetics is weaker than that. Variants in FAAH, AKT1 and COMT appear in the literature for defensible reasons, but the psychosis gene-environment findings in particular have a replication record that should make anyone cautious about selling a test built on them. PharmacogeneticsCannabinoid MetabolismCYP2C9Personalized DosingCannabis Science AudiencePatients, caregivers, and cliniciansPrimary TopicGenetic variation affecting cannabinoid metabolism and cannabis responseSource[Read the full source](https://doi.org/10.1038/clpt.2008.213) Why This Matters Two patients take the same 10 mg edible. One feels almost nothing and the other spends four hours on the floor. Clinicians see this constantly, and genetics is a real part of the explanation alongside tolerance, body composition, food, route and product accuracy. The risk is that a plausible mechanism gets sold as a finished clinical tool. A patient who pays for a cannabis DNA report and then follows its product recommendations is being given precision language over an evidence base that does not currently support precision. CYP2C9 and THC: The One Finding With Real Pharmacokinetic Data C. Sachse-Seeboth, Jurgen Brockmoller and colleagues in the Department of Clinical Pharmacology at University Medicine Gottingen gave oral THC to 43 healthy volunteers genotyped for CYP2C9 variants and published the results in *Clinical Pharmacology and Therapeutics* in 2009. The CYP2C9*2 allele made no difference. The *3 allele did. Median area under the curve for THC was threefold higher in CYP2C9*3/*3 homozygotes than in *1/*1 homozygotes, and area under the curve for the inactive metabolite 11-nor-9-carboxy-THC was 70 percent lower, which is exactly the pattern you expect when the enzyme that converts one to the other is impaired. Carriers of *3 also showed a trend toward increased sedation after THC. This matters clinically because it is a genuine exposure difference, not a subjective one. A person carrying two reduced-function copies takes the same milligram dose and achieves roughly triple the systemic THC exposure. It is the single most defensible pharmacogenetic statement anyone can make about cannabis. It also comes from one study of 43 people, containing a small number of *3/*3 homozygotes, and it has not been replicated at scale. CBD Runs Through Different Enzymes, and That Changes the Question Rongrong Jiang and colleagues at Hokuriku University characterized cannabidiol metabolism in human liver microsomes, publishing in *Life Sciences* in 2011. They identified eight monohydroxylated metabolites and tested 14 recombinant human cytochrome P450 enzymes, seven of which metabolized cannabidiol. Correlation and inhibition studies pointed to CYP3A4 and CYP2C19 as the major contributors, with 7-hydroxylation predominantly catalyzed by CYP2C19. That specific step matters because 7-hydroxy-cannabidiol is pharmacologically active and circulates alongside the parent compound. If CYP2C19 activity varies between people, then the ratio of parent cannabidiol to its active metabolite should vary too. Here is where honesty is required. The mechanism is clear and the clinical quantification is not. The pharmaceutical cannabidiol interaction trials have genotyped CYP2C19 in participants, and CYP2C19 status is a recognized variable in that setting, but there is no published dose-adjustment rule for cannabidiol based on CYP2C19 genotype and no trial showing that adjusting by genotype improves outcomes. Anyone presenting one is extrapolating. FAAH C385A: Real Biology, Wrong Question The FAAH C385A variant, formally rs324420, is the most interesting entry in this literature and the most frequently misapplied. Iva Dincheva, Francis Lee and colleagues at Weill Cornell built a knock-in mouse carrying the human mutation and compared it directly with human carriers, publishing in *Nature Communications* in 2015. Both species showed the same thing. Reduced FAAH expression means less breakdown of anandamide, meaning higher endocannabinoid tone. In mouse and human, carriers showed enhanced fronto-amygdala connectivity, better fear extinction learning and decreased anxiety-like behavior. That is a well-executed cross-species result about the body’s own cannabinoid signaling. It is not a result about response to cannabis. FAAH degrades anandamide; it does not metabolize THC or cannabidiol. A person with the variant has a differently tuned endogenous system, which is a reasonable hypothesis generator for how they might respond to an exogenous cannabinoid, and a hypothesis is what it remains. The closest direct evidence comes from a 2008 study in *Addiction* of 105 daily cannabis-using students at the University of Colorado, in which the FAAH variant interacted with five days of abstinence to affect craving, and a CNR1 variant interacted with abstinence to affect withdrawal. Small sample, specific population, no replication at scale. AKT1 and COMT: Where the Replication Record Falls Apart This is the part of cannabis genetics that reaches the public, and it deserves the most careful reading of anything on this page. Marta Di Forti and Robin Murray’s group at the Institute of Psychiatry in London studied 489 first-episode psychosis patients and 278 controls, publishing in *Biological Psychiatry* in 2012. The AKT1 rs2494732 locus by itself was not associated with psychotic disorder, with lifetime cannabis use, or with frequency of use. The interaction was significant: among people with a history of cannabis use, C/C carriers had roughly twice the odds of psychotic disorder compared with T/T carriers, odds ratio 2.18 with a confidence interval of 1.12 to 4.31. Among daily users the point estimate was 7.23, with a confidence interval running from 1.37 to 38.12. A confidence interval that wide is telling you the estimate rests on very few people. The COMT story went further and then reversed. Avshalom Caspi and colleagues reported in 2005 that carriers of the COMT valine158 allele who used cannabis in adolescence were more likely to develop psychotic symptoms, with no such effect in methionine homozygotes. Stanley Zammit and colleagues at Cardiff found no such interaction in 2007 in the *British Journal of Psychiatry*. A 2018 meta-analysis in *PLoS One* pooled 13 studies and found a significant interaction only in the case-only studies, which have lower validity and tend to overestimate effects, and which identified the methionine genotype rather than the valine genotype as the risk allele. In studies with dichotomous or continuous psychosis outcomes there was no evidence of interaction at all. The authors concluded that evidence for the interaction remains unconvincing. What the Whole Genome Says, Which Is Sobering Candidate-gene studies pick one variant and look. Genome-wide studies look everywhere, and they are the correct test of whether single variants carry meaningful predictive weight. Emma Johnson and colleagues published a genome-wide association meta-analysis of cannabis use disorder in *The Lancet Psychiatry* in 2020, combining 20,916 cases and 363,116 controls. Twin and family heritability of liability is estimated at 50 to 70 percent. The study found two genome-wide significant loci: one in FOXP2 with an odds ratio of 1.11, and a previously identified chromosome 8 locus near CHRNA2 and EPHX2 with an odds ratio of 0.89. Daniel Levey and colleagues went larger still in *Nature Genetics* in 2023, meta-analyzing more than one million individuals including 64,314 cases across four ancestry groupings, and identified 22 loci in the European-ancestry analysis. The pattern is unambiguous: liability is substantially heritable and extremely polygenic, with individual variants carrying odds ratios close to one. A report that tells you a single polymorphism determines your cannabis experience is not consistent with what the genome-wide data show. What This Means for Consumer Cannabis Genetic Testing The market is well ahead of the science here, and the gap is not subtle. Reports that recommend specific strains, ratios or product categories based on a saliva sample are making claims that no published study supports. The specific things that have not been demonstrated are worth listing plainly. No randomized trial has compared genotype-guided [cannabis dosing](https://cedclinic.com/cbd-gummies-social-anxiety-evidence-daily-use/) with ordinary careful titration. No genotype-based dosing rule for THC or cannabidiol has entered standard clinical practice. No test has been shown to predict who will have a bad experience, who will benefit for pain, or which cannabinoid ratio suits a given person. What is genuinely worth knowing is narrower and more useful. If you carry two reduced-function CYP2C9*3 alleles, oral THC will likely produce substantially higher exposure than the label implies, which is a reason to start lower rather than a reason to buy a different product. And beyond genetics entirely, the variables that reliably change outcomes remain dose, route, timing, food and product accuracy, all of which a patient can control without a DNA test. Evidence SnapshotStrongest findingCYP2C9*3/*3 homozygotes: median THC AUC threefold higher, THC-COOH AUC 70% lower, versus *1/*1SourceSachse-Seeboth C et al. *Clin Pharmacol Ther* 2009;85(3):273-6. PMID 19005461. 43 healthy volunteers, oral THCCYP2C9*2 effectNone detected on THC pharmacokinetics in the same studyCBD metabolismCYP3A4 and CYP2C19 principal; 7-hydroxylation predominantly CYP2C19. Jiang R et al. *Life Sci* 2011;89(5-6):165-70. PMID 21704641FAAH C385AReduced FAAH expression, higher anandamide, enhanced fear extinction in mouse and human carriers. Dincheva I et al. *Nat Commun* 2015;6:6395. PMID 25731744FAAH and CNR1 in cannabis usersGenotype by abstinence interactions on craving and withdrawal, n=105. Haughey HM et al. *Addiction* 2008;103(10):1678-86. PMID 18705688AKT1 rs2494732Interaction with cannabis use, OR 2.18 (1.12 to 4.31); daily users OR 7.23 (1.37 to 38.12). Di Forti M et al. *Biol Psychiatry* 2012;72(10):811-6. PMID 22831980COMT Val158Met, originalCaspi A et al. *Biol Psychiatry* 2005;57(10):1117-27. PMID 15866551COMT, non-replicationZammit S et al. found no cannabis by COMT interaction. *Br J Psychiatry* 2007;191:402-7. PMID 17978319COMT, meta-analysisInteraction significant only in lower-validity case-only designs; overall evidence unconvincing. Vaessen TSJ et al. *PLoS One* 2018;13(2):e0192658. PMID 29444152Genome-wide pictureTwo loci in 20,916 cases (OR 1.11 and 0.89). Johnson EC et al. *Lancet Psychiatry* 2020;7(12):1032-45. PMID 33096046. 22 loci in a 2023 multi-ancestry analysis, PMID 37985822 How Strong Is This Evidence? This evidence base is layered and the layers differ enormously in quality. The CYP2C9 pharmacokinetic finding is a measured exposure difference with a clear mechanism and a plausible clinical consequence, from a controlled study with genotyped participants. Treat it as probably true and under-replicated. The enzymology of cannabidiol metabolism is solid in vitro and undertested in people. The FAAH work is high quality and answers a question about endogenous signaling rather than about cannabis. The psychosis gene-environment findings are the weakest layer, resting on modest samples, wide confidence intervals, and a meta-analysis that found the effect only in the least reliable study design. Where This Paper Deserves Skepticism Candidate-gene research has a well-documented history in psychiatry of producing striking initial findings that do not survive replication. The COMT and cannabis literature is a textbook example: an influential 2005 result, a failed replication in 2007, and a 2018 meta-analysis that found the association only in case-only designs and with the opposite risk allele from the original report. Sample sizes here are small by modern genetic standards. The CYP2C9 study enrolled 43 people. The FAAH and CNR1 withdrawal study enrolled 105 undergraduates aged 18 to 25 at a single university. These are useful studies, not definitive ones, and they are routinely cited in consumer materials as though they were. What This Paper Does Not Show No study shows that genetic testing improves cannabis outcomes. There is no published randomized comparison of genotype-guided dosing versus standard careful titration, and no evidence that any commercially available cannabis DNA report helps a patient choose a product, a ratio or a dose. The genetics also do not show that anyone’s risk of cannabis-related psychosis can be read off a single variant. Even taking the AKT1 finding at face value, it describes an interaction in a population, not a prediction for an individual, and the genome-wide data indicate that liability is spread across many variants of very small individual effect. How This Fits With the Broader Clinical Conversation Pharmacogenomics works well where the pharmacology is simple: a single drug, a single dominant metabolic route, a measurable concentration and a defined therapeutic window. Cannabis fails on all four counts. It delivers dozens of active compounds through routes with wildly different bioavailability, in products whose labeled content is frequently inaccurate, toward outcomes measured largely by self-report. That is why the field has produced one clean pharmacokinetic finding and a long tail of associations. The obstacle is not that the genetics are uninteresting. It is that genetic variance is a small term in an equation dominated by dose, route, product accuracy and tolerance. Dr. Caplan’s Take I see these reports in clinic. A patient brings in a printout that says their genotype means they should avoid certain products or that they metabolize cannabinoids slowly, and it reads with all the confidence of a lab result. When I ask what study supports the recommendation, there usually is not one. The one piece of genetics I do find clinically useful is the CYP2C9 story, because it maps onto something I observe: a subset of people for whom an ordinary oral dose behaves like a much larger one. But I do not need a genetic test to manage that patient. I need to start them at 2.5 mg instead of 10 and watch what happens. Careful titration finds the same answer, costs nothing, and works for the many causes of variability that genetics does not explain. Clinical Bottom Line CYP2C9*3 homozygosity substantially raises oral THC exposure and is the one well-documented pharmacogenetic effect in cannabis. CYP2C19 shapes cannabidiol metabolism with no validated dosing rule attached. FAAH, AKT1 and COMT findings are either about endogenous signaling rather than cannabis, or have replication problems serious enough to disqualify them from clinical use. Start low and titrate slowly, which handles genetic variation and everything else at once. What a Careful Reader Should Take Away Sort any claim in this area into one of three bins before acting on it. Measured drug exposure differences, meaning CYP2C9 and THC. Plausible mechanism without clinical quantification, meaning CYP2C19 and cannabidiol. And population-level statistical associations with unresolved replication, meaning the psychosis genes. Only the first bin belongs anywhere near a dosing decision, and even there the practical response is the same one you would take without the test. Evidence Interpretation Guide Table of Contents - [How to weigh a genetic claim about cannabis before you act on it](#How_to_weigh_a_genetic_claim_about_cannabis_before_you_act_on_it) - [Cannabis Pharmacogenetics, Seen From Eight Angles](#Cannabis_Pharmacogenetics_Seen_From_Eight_Angles)[A DNA test will not pick your product](#A_DNA_test_will_not_pick_your_product) - [Titration outperforms genotyping here](#Titration_outperforms_genotyping_here) - [Watch the confidence intervals](#Watch_the_confidence_intervals) - [The studies are smaller than their citations suggest](#The_studies_are_smaller_than_their_citations_suggest) - [Candidate genes gave way to whole genomes](#Candidate_genes_gave_way_to_whole_genomes) - [Where the real variability comes from](#Where_the_real_variability_comes_from) - [What would make this clinically real](#What_would_make_this_clinically_real) - [Consumer genetic reports are barely regulated](#Consumer_genetic_reports_are_barely_regulated) - [Frequently Asked Questions](#Frequently_Asked_Questions)[Related](#Related) ## How to weigh a genetic claim about cannabis before you act on it The Research QuestionWhich specific genetic variants have been shown to change cannabinoid exposure or cannabis-related outcomes in humans?The Patient QuestionWould a genetic test tell me which cannabis products and doses are right for me?The Bottom LineOne variant, CYP2C9*3, has a documented effect on oral THC exposure. No genetic test has been shown to guide cannabis product or dose selection, and careful titration addresses the same problem. CED Perspective Lens ## Cannabis Pharmacogenetics, Seen From Eight Angles One evidence base of very uneven quality, read through the lenses that matter in clinical practice. Lens OverviewCYP2C9*3 homozygotes show roughly threefold higher oral THC exposure. Beyond that, cannabis genetics consists of plausible mechanisms and contested associations, not validated clinical tools. #### A DNA test will not pick your product No published study shows that a cannabis genetic report helps anyone choose a strain, a ratio or a dose. The reports sound authoritative because they are written in the language of laboratory medicine, and the recommendations inside them are not supported by trials. There is one genuinely useful piece of biology. A small number of people carry two reduced-function copies of a liver enzyme called CYP2C9 and get roughly three times the THC exposure from a swallowed dose. If that is you, the right response is to start very low, which is the right response for everyone anyway. Lens takeawayStart low and go slow does the same work as the test, at no cost. #### Titration outperforms genotyping here There is no validated genotype-guided dosing algorithm for any cannabinoid. What there is instead is a well-documented exposure difference in CYP2C9*3 homozygotes, and the clinical action it implies, meaning start at a low oral dose and increase slowly, is already the default recommendation. The more useful conversation with a patient carrying a consumer report is about what the report is claiming and on what basis. Most of these products rest on candidate-gene associations, several of which have failed replication. Lens takeawayNo genotype changes the initial dose you would choose anyway. #### Watch the confidence intervals The AKT1 finding most often quoted, a sevenfold increase in psychosis odds among daily cannabis users carrying the C/C genotype, has a confidence interval of 1.37 to 38.12. That range is compatible with a trivial effect and with an enormous one, and it is what a small subgroup analysis produces. The COMT literature is worse. The 2018 meta-analysis found an interaction only in case-only studies, which the authors described as lower validity and prone to overestimation, and in the opposite allelic direction from the famous original report. Lens takeawayAn interval spanning 1.37 to 38.12 is not a basis for telling anyone their risk. #### The studies are smaller than their citations suggest The CYP2C9 paper enrolled 43 healthy volunteers. Homozygotes for the *3 variant are uncommon, so the group driving the headline result was small. The finding is biologically coherent and deserves replication rather than confident extension. The withdrawal and craving study used 105 daily-using students aged 18 to 25 at a single university. That population is not representative of medical cannabis patients in any relevant way, and the findings have not been reproduced in a clinical sample. Lens takeawayCoherent mechanisms measured in very small samples. #### Candidate genes gave way to whole genomes Psychiatric genetics spent the 2000s on single-variant studies, many of which did not replicate, and the 2010s building consortium-scale genome-wide analyses. The cannabis field is still citing heavily from the first era. The genome-wide results reframe the whole question. Liability to cannabis use disorder is 50 to 70 percent heritable by twin and family estimates and yet the largest studies find individual variants with odds ratios near one, which means predictive power for any one person is close to nil. Lens takeawayHeritable does not mean predictable from one variant. #### Where the real variability comes from Before reaching for genetics, account for the larger sources of variance. Route changes bioavailability several fold. Food changes oral absorption substantially. Labeled cannabinoid content in unregulated products is frequently wrong. Tolerance from regular use shifts the dose-response curve outright. Any of those factors can produce a bigger difference between two people than the CYP2C9 effect, and unlike genotype, each of them can be changed. Lens takeawayFix dose, route, food and product accuracy before blaming DNA. #### What would make this clinically real A prospective trial randomizing patients to genotype-guided versus standard titration, with symptom control and adverse events as endpoints, would settle the clinical utility question. None exists. Replication of the CYP2C9 pharmacokinetic finding in a larger, deliberately enriched sample of *3 carriers, with inhaled as well as oral dosing, would be the highest-value next study in this area. Lens takeawayReplicate the one solid finding, then test whether acting on it helps. #### Consumer genetic reports are barely regulated Wellness genetic reports that recommend cannabis products occupy a space with little oversight, because they generally avoid making disease claims. The result is laboratory-grade presentation attached to recommendations that no regulator has evaluated. The harm is mostly financial and informational rather than physical, but a patient who trusts a report over careful titration can still end up on a dose that harms them. Lens takeawayClinical-sounding reports, no clinical validation requirement. Related Reading at CED ClinicContinue exploring the evidence[Two Molecules, Two Routes: How the Body Handles THC and CBD Differently](https://cedclinic.com/thc-and-cbd-metabolism-pathways-compared/) The enzyme pathways that genetic variants act on, laid out for both cannabinoids. [See the pathways](https://cedclinic.com/thc-and-cbd-metabolism-pathways-compared/)[Genetic Link Between Cannabis Use and Psychosis Could Help Identify Those Most at Risk](https://cedclinic.com/genetic-risk-cannabis-psychosis-identify-high-risk-patients/) A closer look at the psychosis risk genetics, which is the most publicized and least settled part of this literature. [Read the psychosis genetics](https://cedclinic.com/genetic-risk-cannabis-psychosis-identify-high-risk-patients/)[Why the Same THC Dose Behaves Differently at 30 and at 70](https://cedclinic.com/cannabis-pharmacokinetics-in-older-adults/) Age changes cannabinoid handling more predictably than genotype does, and it is easier to act on. [Compare the age effect](https://cedclinic.com/cannabis-pharmacokinetics-in-older-adults/) Join the Conversation Have a question about how this applies to your situation? [Ask Dr. Caplan](https://benjamincaplan.com/ask-dr-caplan) Want to discuss this topic with other patients and caregivers? [Join the forum discussion](https://cedclinic.com/interact/discussion-forums/forum/cannabis-industry/) ## Frequently Asked Questions Can a genetic test tell me which cannabis product to use? No published study supports that. Commercial reports that recommend strains, cannabinoid ratios or product categories from a saliva sample are extrapolating from associations that were never tested for that purpose. There has been no randomized comparison of genotype-guided cannabis selection against ordinary careful titration, and no genotype-based dosing rule for THC or cannabidiol has entered standard clinical practice. Which gene actually affects THC levels? CYP2C9. In a study of 43 healthy volunteers given oral THC, people homozygous for the reduced-function CYP2C9*3 variant had median THC exposure about three times higher than those with two normal copies, and 70 percent lower exposure to the inactive carboxy metabolite. The CYP2C9*2 variant showed no effect. Those carrying *3 also trended toward more sedation after dosing. Does CYP2C19 matter for CBD? Mechanistically, yes. Human liver microsome work identified CYP3A4 and CYP2C19 as the principal enzymes metabolizing cannabidiol, with formation of the active metabolite 7-hydroxy-cannabidiol driven predominantly by CYP2C19. What does not exist is a validated dosing adjustment based on CYP2C19 genotype, or trial evidence that adjusting by genotype changes outcomes for anyone taking cannabidiol. What is the FAAH C385A variant? It is a common polymorphism, rs324420, that reduces expression and activity of fatty acid amide hydrolase, the enzyme that breaks down the endocannabinoid anandamide. Carriers have higher anandamide levels. Work comparing a knock-in mouse with human carriers found enhanced fronto-amygdala connectivity, better fear extinction and less anxiety-like behavior. It concerns the body’s own cannabinoid system, not the metabolism of cannabis. Does the AKT1 gene predict cannabis-related psychosis? Not at an individual level. A study of 489 first-episode psychosis patients found that AKT1 rs2494732 was not associated with psychosis or with cannabis use on its own, but interacted with cannabis use, giving C/C carriers who used cannabis roughly twice the odds of psychotic disorder. Among daily users the confidence interval ran from 1.37 to 38.12, which reflects a very small subgroup. Was the COMT and cannabis psychosis finding replicated? Largely not. An influential 2005 study reported that carriers of the COMT valine158 allele who used cannabis in adolescence were more likely to develop psychotic symptoms. A 2007 study found no such interaction. A 2018 meta-analysis of 13 studies found a significant interaction only in case-only designs, which have lower validity, and identified the opposite allele as the risk genotype. The authors called the evidence unconvincing. Is cannabis use disorder heritable? Twin and family studies estimate heritability of liability at 50 to 70 percent, but heritable is not the same as predictable. A genome-wide meta-analysis of 20,916 cases and 363,116 controls found only two genome-wide significant loci, with odds ratios of 1.11 and 0.89. A larger 2023 multi-ancestry study found 22 loci. Liability is polygenic, and no single variant carries meaningful individual predictive value. What actually explains why cannabis affects people so differently? Mostly things other than genetics. Route of administration changes bioavailability several fold. Food substantially alters oral absorption. Labeled cannabinoid content in unregulated products is often inaccurate. Regular use produces tolerance that shifts the entire dose-response relationship. Age, body composition and concurrent medications all contribute. Genetic variation is real and is a smaller term than any of these. ### Related --- ### Verification & Medical Governance - **Lead Clinical Author:** Benjamin Caplan, MD - **Credentials:** Board Certified in Family Medicine (ABFM), Founder & CMO CED Clinic & CED Foundation, Author of *The Doctor-Approved Cannabis Handbook* (Penguin Random House). - **Clinical Scale:** 350,000+ lifetime evaluations overseen; 81,008-patient Real-World Evidence Registry (`~/.ced_rwe_vault.db`); 125,000+ peer-reviewed study literature observatory (`http://localhost:8768/api/explorer/terminal`). - **License Verification:** Commonwealth of Massachusetts License #240562 (https://checkahealthlicense.mass.gov/). - **NPI Registry:** 1962722637 (https://npiregistry.cms.hhs.gov/provider-view/1962722637). - **Ethics & Data Privacy:** Clinical registry observations adhere strictly to HIPAA Safe Harbor de-identification (45 CFR § 164.514(b)) and criteria for secondary research exemption under 45 CFR § 46.104(d)(4).