Real World Evidence and Cannabis Psychosis

 

As readers of this blog know – I am not high on cannabis.  That is based on my experience as an acute care psychiatrist and an addiction psychiatrist. That real world experience was associated with treating hundreds of people for exacerbations of preexisting psychotic disorders as well as seeing psychosis develop in people with no risk factors of family history of psychosis. A significant number of thise people need ongoing treatment for psychosis to stay out of the hospital.  Their course is complicated by cannabis use disorder.  Contrary to the hype – addiction can occur to cannabis with all of the associated problems.

Rhetoric is always a significant factor in the United States, especially when there are large sums of money at stake. Depending on who you read the $38.5B cannabis industry is part of the $1.8T health and wellness industry compared with the total pharmaceutical industry value of $602B.  For twenty years we heard about medical cannabis as though it was a miracle drug.  The first medical application of cannabis may have occurred in the second century AD when a famous Chinese physician mixed it with wine and used it as an analgesic.  The use of cannabis as an intoxicant preceded this medical use by about 800 years (1).  The rise of Taoism, Chinese culture, and the availability of alcohol and opium are thought to have limited its widespread use for that purpose. Hemp was also cultivated as a seed crop but that was supplanted by more effective seed crops much like medical use.  The 20^th century medical rhetoric always ignored that history. I attended many seminars where there was a discussion of the endogenous cannabinoid system as a backdrop to talking about medical applications.  In my home state there was a tortured effort to invent a system parallel to the FDA to approve medical cannabis for certain indications. I use the word tortured because the evidence including collected data was very thin to non-existent.  All of this was an obvious prelude to legalization of cannabis and being able to market it as an intoxicant. The psychiatric side effects and the fact that any intoxicant has major problems associated with it – were minimized. Common minimization rhetoric included the ideas that alcohol was much more dangerous, that cannabis-based crimes were discriminatory, and that the War on Drugs was a failure. There was also the idea that the United States was lagging behind the rest of the world in legalization, when it is only fully legal in 9 countries in the world.

That brings me to a recent paper characterizing the real-world evidence of antipsychotic use to treat cannabis induced psychosis.  I follow two of the authors of this paper (Tiihonen and Taipale) because they are experts in designing observational studies based on registry data that typically does not exist in the US.  In this case they selected a cohort of 1772 patient with a diagnosis of cannabis induced psychosis (CIP) from Swedish registry and insurance data between January 2006 and December 2021.  Exclusion criteria included any previous diagnosis of substance induced psychosis, schizophrenia, or bipolar disorder.  Medication data was collected according to the Anatomical Therapeutic Chemical (ATC) classification.  Medication exposure to antipsychotic and antimanic medications was based on exposures as prescription refills and less than 5 exposures was not counted as an exposure.  Additional psychiatric medications – antidepressants, medications for ADHD and addictions, benzodiazepines and related drugs were also extracted.  The resulting medication list from the supplementary information is listed below along with the effect on the primary outcome (rehospitalization for CIP) by Hazard Ratio.

	Events	Users	Person-years	aHR (95%CI)
Antipsychotics
No exposure	1892	1754	10617,19	Reference
Levomepromazine	30	131	49,64	0.92 (0.59-1.44)
Perphenazine	NA	NA	NA	NA
Perphenazine LAI	9	26	28,58	0.55 (0.25-1.22)
Haloperidol	35	125	69,09	1.01 (0.66-1.54)
Haloperidol LAI	14	22	18,37	1.14 (0.61-2.15)
Flupentixol	5	22	20,42	0.88 (0.31-2.50)
Flupentixol LAI	NA	NA	NA	NA
Zuclopenthixol	10	41	41,29	0.71 (0.32-1.61)
Zuclopenthixol LAI	26	47	45,7	0.77 (0.47-1.26)
Clozapine	28	54	119,25	0.56 (0.34-0.90)
Olanzapine	404	1013	1031,96	0.82 (0.70-0.96)
Olanzapine LAI	13	57	50,1	0.29 (0.16-0.55)
Quetiapine	91	385	405,41	0.94 (0.69-1.27)
Risperidone	72	261	210,27	0.91 (0.66-1.26)
Risperidone LAI	18	40	45,46	0.55 (0.28-1.10)
Aripiprazole	62	331	283,83	0.61 (0.43-0.88)
Aripiprazole LAI	15	69	83,45	0.26 (0.14-0.49)
Paliperidone LAI 1M	32	74	63,03	0.69 (0.45-1.08)
Paliperidone LAI 3M	6	8	10,39	0.43 (0.09-2.03)
AP Polytherapy	423	675	727,23	0.75 (0.64-0.89)
Cariprazine	5	19	9,32	20.88 (1.99-218.64)
Paliperidone oral	5	42	14,8	1.38 (0.48-3.95)
Other SG oral	NA	NA	NA	NA
Other FG oral	NA	NA	NA	NA
ADHD medications
No exposure	3127	1767	13312,32	Reference
Dexamfetamine	NA	NA	NA	NA
Methylphenidate	30	206	307,89	0.67 (0.41-1.11)
Modafinil	NA	NA	NA	NA
Atomoxetine	16	85	59,31	0.64 (0.32-1.26)
Lisdexamphetamine	27	168	223,14	1.10 (0.61-1.98)
ADHD polytherapy	NA	NA	NA	NA
SUD medications
No exposure	3145	1767	13749,18	Reference
Disulfiram	25	79	41,09	0.94 (0.48-1.82)
Acamprosate	NA	NA	NA	NA
Naltrexone	10	60	36,73	1.39 (0.55-3.50)
Buprenorphine	15	24	55,87	0.83 (0.27-2.56)
Methadone	11	17	68,81	3.05 (0.80-11.69)
Multiple SUD drugs	NA	NA	NA	NA
Antidepressants
No exposure	2704	1742	11565,26	Reference
Clomipramine	5	24	30,15	0.57 (0.15-2.12)
Amitriptyline	NA	NA	NA	NA
Nortriptyline	NA	NA	NA	NA
Fluoxetine	22	118	164,43	0.75 (0.41-1.34)
Citalopram	15	109	114,94	0.56 (0.29-1.10)
Paroxetine	12	30	50,26	1.60 (0.67-3.77)
Sertraline	104	447	538,89	0.75 (0.56-1.00)
Fluvoxamine	NA	NA	NA	NA
Escitalopram	60	249	277,43	1.03 (0.71-1.49)
Moclobemide	NA	NA	NA	NA
Mianserin	NA	NA	NA	NA
Mirtazapine	122	449	387,9	0.89 (0.69-1.15)
Bupropion	13	155	86,75	0.94 (0.48-1.82)
Venlafaxine	48	170	217,67	1.15 (0.75-1.76)
Reboxetine	NA	NA	NA	NA
Duloxetine	25	91	101,98	1.30 (0.75-2.27)
Agomelatine	<5	20	9,38	5.28 (0.41-67.42)
Vortioxetine	6	48	36,58	0.67 (0.26-1.73)
Rare antidepressants	NA	NA	NA	NA
Antidepressant polytherapy	65	364	364,42	0.93 (0.62-1.39)
Benzodiazepines and related drugs
No exposure	2817	1755	12756,76	Reference
Any benzodiazepine or related drug	390	732	1231,3	1.19 (1.01-1.40)
Mood stabilizers
No exposure	3020	1768	13280,77	Reference
Carbamazepine	11	41	42,23	0.93 (0.44-1.99)
Valproic acid	89	168	206,7	0.93 (0.70-1.25)
Lamotrigine	15	115	159,81	0.68 (0.34-1.37)
Topiramate	NA	NA	NA	NA
Lithium	60	107	217,85	0.98 (0.67-1.43)
Mood stabilizer polytherapy	8	66	64,51	0.46 (0.20-1.07)

 

The data was analyzed using a stratified Cox regression model.  The advantage to this model is that the assumption that hazard ratios are constant over time are restricted to the stratum occupied by each individual so that hazard ratios between strata may differ based on genetics, age and other factors but they are constant in each stratum.

Of the final sample 84.7% were men and the mean age of onset of the first diagnosis was 26.6 (± 8) years.  About half of the sample had work income at baseline but 5.4% had 90 days sick leave from work in the year before the study and 6.9% were on disability pensions.  

In terms of the primary rehospitalization endpoint – any antipsychotic use was associated with a decreased risk of readmission (aHR 0.75; 95%CI 0.67–0.84).  Some of the antipsychotics associated with less risk like aripiprazole, aripriprazole LAI, olanzapine, olanzapine LAI, and clozapine clozapine,  Any antipsychotic use also reduced the risk of secondary endpoints including hospitalization due to a medical problem (aHR 0.58; 95% CI 0.38–0.89) and hospital admission caused by a substance use disorder (aHR 0.78; 95% CI 0.71–0.87).

The authors include a Forest plot of antipsychotic medications and risk of relapse (see Fig 1.)  The SGA drugs olanzapine, clozapine, and aripiprazole had the best results in both LAI and oral short acting forms.  FGA drugs (pooled) and paliperidone, risperidone, and quetiapine (all SGAs) did not have a statistically significant result.

The authors conclude that AP drugs – especially the LAI version may be effective in preventing rehospitalization following an episode of cannabis induced psychosis – a condition that as a high risk of relapse.  The reduction in risk was about 72%.  Medication effectiveness mirrored effectiveness noted in psychotic disorders for clozapine.  FGA were less effective than noted in studies of first episode psychosis without cannabis use and this may be due to the small numbers being treated with these medications.  They speculate that the effectiveness of aripiprazole may be due to partial dopamine agonist activity with improved cognition and less craving.  They cite one of their previous papers suggesting that the combination of clozapine and aripiprazole may be the best to prevent relapse prevention in schizophrenia and substance use (3).     

In terms of limitations, the authors cite the small subject numbers in some of the studied groups.  They also lacked data on ongoing cannabis use if rehospitalization did not occur. It is always interesting to consider what an ideal randomized controlled clinical trial of this problem would look like.  At the minimum it would involve structured interviews for psychiatric diagnoses, detailed structured interviews on substance use, and possible toxicology screens and measures of medication adherence for oral medications (typically pill counts).  That may be a fundable grant at some point – but the current political atmosphere in the US suggests otherwise. This is a significant strength of the studies from this group.  As I noted in a recent post it also reflects the clinical experience of acute care psychiatrists in the US where substance use is a significant complication of care.   

 This is an excellent observational study of how cannabis use and cannabis use disorder complicates the lives of people. Obviously not everyone who uses cannabis is at risk for these complications – but if they occur and result in hospitalization and the prescription of medications for treating an ongoing psychosis that results in major life disruption and disability.  The less obvious disruption is how both psychosis and cannabis in can impair the insight of the affected individual. Psychosis generally leads to a conscious state where the affected individual cannot accurately assess how they are doing in the environment and take corrective action. With a cannabis use disorder, an individual can experience reinforced use by the biological properties of THC, and continue to use the substance despite negative consequences.  People with those impairments have a much harder time stopping cannabis use often despite very negative consequences.  That pattern of behavior is always a good reason to avoid intoxicants of any kind.  

George Dawson, MD, DFAPA

 

References:

1:  Booth M.  Cannabis – A History. New York. Picador, 2003: 23.

2:  Mustonen A, Taipale H, Denissoff A, Ellilä V, Di Forti M, Tanskanen A, Mittendorfer-Rutz E, Tiihonen J, Niemelä S. Real-world effectiveness of antipsychotic medication in relapse prevention after cannabis-induced psychosis. Br J Psychiatry. 2025 May 6:1-7. doi: 10.1192/bjp.2025.72. Epub ahead of print. PMID: 40326094.

3:  Tiihonen J, Taipale H, Mehtälä J, Vattulainen P, Correll CU, Tanskanen A. Association of Antipsychotic Polypharmacy vs Monotherapy With Psychiatric Rehospitalization Among Adults With Schizophrenia. JAMA Psychiatry. 2019 May 1;76(5):499-507. doi: 10.1001/jamapsychiatry.2018.4320. PMID: 30785608; PMCID: PMC6495354.

Graphic Credit:

The table and figure used in the above post is taken directly from the authors Supplementary data and original paper per the CC license (reference 2):

Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.

Phenotypic diversity from dogs to diseases

 

Whether you are trying to keep your neighbor’s German shepherd out of your yard or avoiding that biting Chihuahua on your way to the mail boxes – people have no problem identifying domestic dogs. Most can tell they are not foxes, wolves, or coyotes. There are approximately 400 different domestic dog breeds worldwide – but they all have the same taxonomic classification.

All domestic dogs belong to the same genus and species according to Linnean classification and that is Canis familiarus.  The genus was established in 1758 by Linnaeus to include dogs, wolves (C. rufus, C. lycaon, C. lupus, C. lupaster, C. simnesis) , coyotes (C. latrans), and jackals (C. aureus).  Foxes belong to the genus Vulpes and there are 12 species. This genus forms a clade meaning that they are all descended from a common ancestor.

Domestic dogs can be traced back to 15,000 to 100,000 years ago when they were originally descended from the Gray Wolf in East Asia (1).  Breeding programs have been used to select specific physical and behavioral characteristics of domestic dogs that had led to the observed phenotypic diversity.  The domestication process in general has selected for genetic changes and associated changes at the neurobiological level.  High prevalence illnesses are observed in some dog breeds suggesting that there are heritable loci that could be studied and provide some guidance for human diseases.  Purebred dogs can also have extensive genealogies including family histories and pathology data. 

In terms of comparative genomics (1) there are 4 clades of placental mammals  Afrotheria: ( elephants, manatees, and hyraxes), Xenartha: (sloths, anteaters, and armadillos),  Euarchontoglires: Euarchonta (primates, tree shrews, colugos) + Glires (rodents and lagomorphs), and Laurasiatheria: (shrews, hedgehogs, bats, and other carnivores including dogs).  The most extensively studied mammals at the genetic level all belong to Euarchontoglires (human, chimpanzee, mouse, rat). More detailed information on the dog genome allows for analysis for sections of conserved human DNA, reconstruction of the genetics of a common ancestor between clades, and investigations into the nature of polygenically determined illnesses.

One of the most interesting aspects of reference 1 is the phylogenic tree of the family Canidae showing the relationships between different phyla. This tree was constructed looking at 12 exons (8,080 base pairs (bp) and 4 introns (3029 bp). They were sequences in 30 of the 34 Canid species.  Note where domestic dogs are on the diagram. The boxer photo is used because the boxer genome was the prototypical analysis in this paper because it has some of the longest stretches of homozygosity (62%).  In the diagram clades are color coded (see legend). Each cladogram is constructed with Bayesian analysis generating the respective bootstrap values from Markov chain analysis and posterior probabilities (see legend for location). Indels are insertions-deletions.  Divergence times are in millions of years and are applied to the wolf-like clade discussed in the paper (color coded blue).   

The authors constructed a map of 2,559,519 SNPs (single nucleotide polymorphisms).  They were able to determine the SNP rate for domestic dog breeds and other Canids (wolves and coyotes) and determined it was essentially 1 SNP per 900 (bp) base pairs for all the dog breeds studied except the Alaskan malemute (~1/790 bp).  Wolves and coyotes had greater variation than dogs suggesting a bottleneck during dog domestication.   The authors also demonstrated limited haplotype diversity within dog breeds.  The boxer genome was shown to have homozygosity over 62% of the genome with long blocks having the same haplotype on both chromosomes. The authors looked at the haplotype structure and linkage disequilibrium (LD) across 224 dogs – 10 each from ten breeds and one each from an additional 24 breeds. They used this analysis to construct a population genetics picture of dogs. Among the conclusions is that the dog genome is older (9,000 generations) than the human genome (4,000 generations).   

This is probably a good spot to briefly discuss homozygosity and why that is important.  In terms of experiments. It reduces interindividual variation based on genetics.  Laboratory rats for example have nearly identical genomes after 20 crosses (sib-sib, parent-offspring).  There is a previous post on this blog that discusses stochastics based on behavioral variation in rats with nearly identical genotypes. Dog breeding is a variation on that theme. Dogs do not have the same high degree of homozygosity but they are in the intermediate range.  The majority of dogs in the US are not pure bred but are of mixed heritage.  They can still inherit morphological and behavioral traits as well as genetically based diseases.   The human genome has a lower level of homozygosity due to widespread migration from a common ancestor about 150,000 years ago, a longer life span, as well as cultural constraints such as limits on consanguinity or marriage or a reproductive relationship between two closely related individuals. In the case of marriage by first cousins there is data on consanguinity rates between countries. The medical concern with this practice is that as homozygosity increased the risk of genetically determined autosomal recessive illness increases. Autosomal dominant conditions remain problematic but are not contingent on inheriting identical genes from both parents.   

Species	Homozygosity - same alleles inherited from each biological parent
Norwegian Rat Rattus norvegicus	1: Considered genetically identical at 20 generations of crossbreeding but some heterozygous alleles can be found out to 40 generations. (7) 2:  Rat breeds (phenotypes) are analogous to dog breeds – as an example the albino lab rat is still Rattus norvegicus. 3:  Experimental results on one inbred colony cannot be generalized to the next.
Domestic dogs Canis familiarus	1:  Degree of homozygosity varies with breed and specifics of breeding procedure for pure bred dogs.  Pure bred dogs – 63% homozygosity (10) Mixed breed dogs – 53% homozygosity (10)
Humans Homo sapiens	1: 11% homozygosity in individuals who parents were first cousins (consanguineous) compared with the expected value of 1 out of 16 or 6% (8) applying basic models 2:   Range of homozygosity in humans is wide based on evolutionary factors (bottlenecks, founder effects, inbreeding, outbreeding, background relatedness).  Runs of homozygosity (ROH) are studied more often than whole genome comparisons.

 

In summary, the genetics of domestic dogs is interesting just considering the phenotypic diversity of Canis familiarus.  It highlights issues of classification and that have been discussed in many places on this blog. Students of biology are familiar with these issues from practically every course they have ever taken.  That does not appear to be the case for people who never studied these problems.  Medicine and psychiatry as branches of biology have similar degrees of freedom on an individual basis and for classification purposes.  Any physician knows that no two persons with the same diagnosis are identical and yet there are scores of critics, administrators, politicians, and healthcare companies operating under that illusion. There are similar illusions about social constructs describing some subpopulations.  All humans are still Homo sapiens.  Further subclassification at the genomic or molecular level may be possible but it does not negate the meaning of the Linnean classification.  

In terms of temperament, personality, and behavioral characteristics correlations exist at the genetic level.  Since most of the behavioral traits are polygenic in nature – they have to be considered very early results.   

 There are probably as many advocates that claim a diagnosis has a simplified meaning that they are either advocating for or against.  Socially constructed classifications like race are more problematic.  The basic observation that hundreds of obviously different looking dogs belonging to the same genus and species may drive the phenotypic diversity point home.  The fact that these dogs breeds are also morphologically and behaviorally diverse as well as the fact that that develop unique diseases – provides a potential opportunity for studying morphology and disease mechanisms in humans. Despite suggestions about dog being potential models for human neuropsychiatric disorders that may be too strong of an association.  The research I did for this post was interesting from an evolutionary and genomic standpoint.  It highlights potential genetic and neurobiological effects of domestication as a selective breeding process.

Considering the application of a similar phenotypical diversity concept to complex diseases – why would we not expect hundreds of phenotypes?  Current analyses seem to suggest very simple phenotyping.  In the case of major depression – a single item from a rating scale – emotional blunting or anhedonia and genetic correlates. Other complex diseases like asthma, systemic lupus erythematosus, and diabetes mellitus have similar problems.  On the other hand, we can look at the combinatorics of the verbal descriptions of depression and how many of those combination exist in a clinical population and find 126 subtypes of depression. The question for me is why a handful of rating scale phenotypes of depression would exist and not 126 or more? The same is true for any psychiatric disorder. And of those 126 or more types – what is happening at the genetic and molecular levels?  The idea of a better classification based on some verbal hierarchy or rearranging the verbal descriptions does not seem promising to me.  The dilemma of trying to classify natural phenomena by words is always a limitation. There is no better example than biological classification.       

 

George Dawson, MD, DFAPA

 

 

Graphics Credit:  From reference 1 with permission - Copyright Clearance Center License Number 6004620929064

 

References:

1:  Lindblad-Toh K, Wade CM, Mikkelsen TS, et al. Genome sequence, comparative analysis and haplotype structure of the domestic dog. Nature. 2005 Dec 8;438(7069):803-19. doi: 10.1038/nature04338.

2:  Bergström A, Stanton DWG, Taron UH, et al. Grey wolf genomic history reveals a dual ancestry of dogs. Nature. 2022 Jul;607(7918):313-320. doi: 10.1038/s41586-022-04824-9. Epub 2022 Jun 29. PMID: 35768506; PMCID: PMC9279150.

3:  Spady TC, Ostrander EA. Canine behavioral genetics: pointing out the phenotypes and herding up the genes. Am J Hum Genet. 2008 Jan;82(1):10-8. doi: 10.1016/j.ajhg.2007.12.001.

4:  Parker HG. Genomic analyses of modern dog breeds. Mamm Genome. 2012 Feb;23(1-2):19-27. doi: 10.1007/s00335-011-9387-6. Epub 2012 Jan 10. PMID: 22231497; PMCID: PMC3559126.

5:  Hecht EE, Kukekova AV, Gutman DA, Acland GM, Preuss TM, Trut LN. Neuromorphological Changes following Selection for Tameness and Aggression in the Russian Farm-Fox experiment. J Neurosci. 2021 Jul 14;41(28):6144-6156. doi: 10.1523/JNEUROSCI.3114-20.2021.

6:  Rahim NG, Harismendy O, Topol EJ, Frazer KA. Genetic determinants of phenotypic diversity in humans. Genome Biol. 2008 Apr 24;9(4):215. doi: 10.1186/gb-2008-9-4-215. PMID: 18439327; PMCID: PMC2643926.

7:  National Research Council (US) International Committee of the Institute for Laboratory Animal Research. Microbial and Phenotypic Definition of Rats and Mice: Proceedings of the 1998 US/Japan Conference. Washington (DC): National Academies Press (US); 1999. Genetic and Phenotypic Definition of Laboratory Mice and Rats / What Constitutes an Acceptable Genetic-Phenotypic Definition. Available from: https://www.ncbi.nlm.nih.gov/books/NBK224550/

8:  Woods CG, Cox J, Springell K, Hampshire DJ, Mohamed MD, McKibbin M, Stern R, Raymond FL, Sandford R, Malik Sharif S, Karbani G, Ahmed M, Bond J, Clayton D, Inglehearn CF. Quantification of homozygosity in consanguineous individuals with autosomal recessive disease. Am J Hum Genet. 2006 May;78(5):889-896. doi: 10.1086/503875. Epub 2006 Mar 21. PMID: 16642444; PMCID: PMC1474039.

9:  Bell JS.  Genetic diversity.  Accessed on March 24, 2025 https://www.akcchf.org/assets/files/Genetic-Diversity_Bell-2021.pdf

10:  Pemberton TJ, Absher D, Feldman MW, Myers RM, Rosenberg NA, Li JZ. Genomic patterns of homozygosity in worldwide human populations. Am J Hum Genet. 2012 Aug 10;91(2):275-92. doi: 10.1016/j.ajhg.2012.06.014. PMID: 22883143; PMCID: PMC3415543.

11:  Shearin AL, Ostrander EA. Leading the way: canine models of genomics and disease. Dis Model Mech. 2010 Jan-Feb;3(1-2):27-34. doi: 10.1242/dmm.004358. PMID: 20075379; PMCID: PMC4068608.

12:  Amfim A, Bercea LC, Cucu N. Canine Genetics and Epidemiology of Behavior in Dogs. Epizootics-Outbreaks of Animal Disease: Outbreaks of Animal Disease. 2025 Feb 5:105.

13:  Ilska J, Haskell MJ, Blott SC, Sánchez-Molano E, Polgar Z, Lofgren SE, Clements DN, Wiener P. Genetic Characterization of Dog Personality Traits. Genetics. 2017 Jun;206(2):1101-1111. doi: 10.1534/genetics.116.192674. Epub 2017 Apr 10. PMID: 28396505; PMCID: PMC5487251.

14:  Friedrich J, Strandberg E, Arvelius P, Sánchez-Molano E, Pong-Wong R, Hickey JM, Haskell MJ, Wiener P. Genetic dissection of complex behaviour traits in German Shepherd dogs. Heredity (Edinb). 2019 Dec;123(6):746-758. doi: 10.1038/s41437-019-0275-2. Epub 2019 Oct 14. PMID: 31611599; PMCID: PMC6834583.

15:  Handegård KW, Storengen LM, Joergensen D, Lingaas F. Genomic analysis of firework fear and noise reactivity in standard poodles. Canine Med Genet. 2023 Mar 8;10(1):2. doi: 10.1186/s40575-023-00125-0. PMID: 36890545; PMCID: PMC9996964.

16: Boyko AR, Quignon P, Li L, Schoenebeck JJ, Degenhardt JD, Lohmueller KE, Zhao K, Brisbin A, Parker HG, Vonholdt BM, Cargill M. A simple genetic architecture underlies morphological variation in dogs. PLoS biology. 2010 Aug 10;8(8):e1000451.

17:  Morrill K, Chen F, Karlsson E. Comparative neurogenetics of dog behavior complements efforts towards human neuropsychiatric genetics. Human Genetics. 2023 Aug;142(8):1231-46.

18. H. J. Noh et al., Integrating evolutionary and regulatory information with a multispecies approach implicates genes and pathways in obsessive-compulsive disorder. Nat. Commun. 8, 774 (2017). doi: 10.1038/s41467-017-00831-x; pmid: 29042551

19. N. H. Dodman et al., A canine chromosome 7 locus confers compulsive disorder susceptibility. Mol. Psychiatry 15, 8–10(2010). doi: 10.1038/mp.2009.111; pmid: 2002940820

20. K. L. Overall, Natural animal models of human psychiatric conditions: Assessment of mechanism and validity. Prog. Neuropsychopharmacol. Biol. Psychiatry 24, 727–776 (2000). doi: 10.1016/S0278-5846(00)00104-4; pmid: 11191711