The present study examined the bivariate and multivariate relationships of the principal components of five chromosomal pathologies (trisomies 13, 18, 21, the monosomy Turner syndrome and Deletion 22q11.2) and five cancers (thyroid, liver, breast and pancreatic cancer and acute myeloid leukaemia) to substance exposure, income and ethnicity covariates in a quantitative causal analysis framework. The main results of the present study were that all ten pathologies examined are rising both across time and in relation to the five cannabinoids examined [cannabis, THC, cannabinol (CBN), cannabichromene (CBC) and cannabigerol (CBG)] in a manner which was robust to adjustment in multivariable inverse probability weighted instrumental variable and robust generalized regression models. Cannabis and all four cannabinoids were significantly related to the PC1’s for chromosomal anomaly rates both before and after adjustment for ETOPFA’s and to the cancer PC1 in final regression models after full adjustment for sociodemographic factors. Large effect sizes were demonstrated for both congenital anomaly PC’s between the highest quintile of cannabis use and the third and fourth quintiles and a medium effect size was shown for the cancer PC. Large effect sizes were also demonstrated between states with legal and not legal cannabis legal regimes for chromosomal anomaly PC’s and a medium effect size was demonstrated between illegal and liberal legal paradigms for the cancer PC1. These effect sizes were accompanied by appropriately large minimal E-values of greater than five and 2.4 respectively, and small P-values (P = 9.67 × 10^–43 and P = 2.66 × 10^–15 respectively).

Previous studies have reported individual conditions in the young known to be associated with cannabis-related chromosomal damage: current data highlight cannabis damage on multiple human chromosomes. Live-born congenital anomaly rates are known to underestimate true rates due to both spontaneous and induced abortion of damaged foetuses. The existence of an experimentally well described threshold dose in the micromolar range^73–81 explains both why inherited morbidity is not observed more commonly at present after prenatal exposure and why rates jump abruptly with the increased use, availability and concentration implicit under the legalization paradigm as shown clearly by epidemiological data from Colorado, Canada, Hawaii and Australia^{8,10,37,82,83}.

It is intriguing to note the variety of chromosomal-toxic mechanisms which are implied by the present results. Whereas the trisomy/monosomy (Turner) syndromes are presumably related to chromosomal mis-segregation errors⁴, horizontal transversions and gene amplifications on chromosome 12 in testicular cancer^{43–46,84,85} and deletion of the short arm of chromosome 22 in Deletion 22q11.2 signify multiple pathways to major chromosomal pathology.

This report studied five congenital chromosomal anomalies and five cancers. Dose–response effects were noted for many associations which were independently significant for cannabis, THC, cannabichromene and cannabigerol and these relationships were preserved after adjustment for estimated ETOPFA rates which is a preponderant effect especially for chromosomal anomalies which are terminated prior to birth at high rates. All ten disorders were noted to rise strongly over time and in relationship to cannabis and the cannabinoids THC, cannabigerol, cannabinol and cannabichromene. Strong effects by quintile of cannabis exposure were noted which were also reflected in the impact of cannabis legal status. Many effect sizes were noted to be strong with many Cohen’s D’s above 1.4, where relationships above 0.8 being typically described as being strong⁸⁶. The pooled disorder-cannabinoid relationship satisfied formal criteria of causality as assessed by inverse probability weighting of robust marginal structural models with P-values significant from 2.8 × 10^–7 and 55/56 e-Values being greater than 1.25 which is the causal threshold and 30/56 minimum e-Values being greater than 5 which is relatively large. The individual disorders are also the subject of separate space–time and causal inference analyses which are presently being prepared.

The findings are prominent for showing a dramatic rise from the fourth to the fifth quintile of cannabis use and a reflection of a similar kind when considering the cannabis legal paradigm. This is reminiscent of a similar finding for 34 defects recently published which also showed a major jump from the fourth to fifth quintile³⁹. This is concerning because it directly reflects the well described exponential dose response which has been found in many cannabinoid genotoxicity studies and in many studies of the pharmacology of cannabinoids generally^{14,26,59–61}.

A direct corollary of this bench to bedside parallelism is that as the community moves steadily into higher echelons of cannabis use the genotoxic sequelae will be unprecedentedly magnified—in coming generations. It is this multi-generational and transgenerational aspect of cannabinoid genotoxicity which is of particular concern in the context of disproportionate dose-exposure escalation.

Cannabis use amongst young adults has been unanimously linked in four of four studies with the subsequent development of testicular cancer^43–46. Testicular cancer is interesting in that 90% of cases involve the formation of an isochromosome 12, and in the remainder an internal intra-chromosomal amplification of parts of the long arm of chromosome 12 occurs so that the relative gene dosage is increased under both scenarios⁴⁰. Cannabis exposure—testicular oncogenesis dose–response effects have been described in several epidemiological series^43–45. Testicular cancer is believed to arise from pro-oncogenic germ stem cell mutations which occur during in utero life which are subsequently activated by the hormonal surge of pubertal development^40–42. In the case of this tumour therefore cannabis accelerates the subclinical pro-oncogenic phase from several decades to just a few years.

Most particularly, the present demonstration of cannabinoid-linked genotoxicity applying to over 500 MB of the human genome accommodated on chromosomes 12, 13, 18, 21 and X clearly indicates that mechanisms exist in man linking in vitro genotoxic effects to clinical effects. Hence it becomes plausible to link the 21 congenital defects noted in Hawaii with cannabis-only exposure³⁷, the 13 congenital anomalies noted in Australia³⁸, limb defects noted in France and Germany^87,88, 29% increase in total congenital anomalies listed in Colorado⁸ a tripling of total birth defects in the high-cannabis using areas of Canada¹⁰ and 34 congenital anomalies in USA³⁹ to transgenerational cannabinoid-induced genotoxic mechanisms. Evidence presented herein also implicates the PC1 of five cancers seen clinically including acute myeloid leukaemia which has been previously documented^89,90 and breast cancer which is the most common cancer of all across USA with 279,100 cases expected in 2020⁹¹. Cannabis has previously been linked with both the induction and promotion of liver fibrosis and cirrhosis^92,93 and with hepatic neocarcinogenesis by numerous mechanisms⁹⁴. It has not been previously linked with the other cancers studied to our knowledge.

Central to any discussion of genotoxic mechanisms of cannabis are considerations of the biological mechanisms by which it mediates chromosomal derangements and disruptions. Cannabis and the cannabinoids THC, cannabinol, cannabidiol and cannabinol have been shown to be toxic to oocytes⁶, sperm^24,95,96, chromosomes⁹⁶, the bases of DNA⁹⁷ and epigenetic regulation both by DNA methylation^{4,19,20,22–24,98} and histone formation^16,99. Cannabidiol and cannabidivarin in low doses have been shown to directly oxidize DNA bases which is a highly oncogenic and mutagenic action⁹⁷. Cannabinoids have long been recognized to reduce the synthesis of major macromolecules of life including DNA, RNA proteins and histones^{13,18,60,99–104}. Reduction in the linker histone H1 has recently been shown to comprise a major oncogenic mechanism by making genes more accessible for transcription¹⁰⁵. One of the proteins whose synthesis is impeded is tubulin^4,16. Tubulin polymerization has many key roles in side the cell including the formation of the microtubules of the mitotic spindle and the molecular skeleton of axons, cilia, centrosomes and flagella¹⁰⁶. Deranged microtubular function has been linked with chromosomes sliding off the mitotic spindle in anaphase and the formation of micronuclei^4,6,7,107 which are described as being a major generator of the genetic chaos of cancer^4,107–118. Indeed, just as histones undergo post-translational modifications tubulin has also been shown to undergo post-translational modifications which target the tubulin monomers for different subcellular destinations¹⁰⁵. Errors in this “tubulin code” have been linked to disorders of flagellar function so that sperm are not able to swim normally in a linear trajectory and go round in circles and fail to correctly target oocytes which is believed to be a potentially significant cause of male infertility¹⁰⁵. Hence whilst for descriptive purposes it is useful to describe cannabinoid-related molecular aberrations in various stratified layers and subcellular compartments, it seems likely that in reality the various layers are intimately crosslinked and molecularly interdependent^{53,54,119–121}. Unfortunately space precludes a more detailed discussion in this forum but many of the important issues have been addressed elsewhere^{4,8–12,14,15,19,20,23,24,38,39,61,102,122,123}.

We feel that our results are widely generalizable for several reasons. The cannabinoid-genotoxicity relationship fulfils most of the Hill criteria for causality including of strength of association, consistency among studies, specificity, temporality, coherence with known data, biological plausibility, dose–response relationship, analogy with situations internationally and a rich experimental research base¹²⁴. Naturally further details will be provided in manuscripts addressing each of these issues individually including consideration in the native space–time data context and formal causal inferential analyses. As noted many of the above findings have been replicated several times elsewhere particularly with relation to the congenital chromosomal anomalies^8,10,37,38. These results are based on the best data available globally. Since causal relationships were demonstrated herein we would expect these relationships to be maintained wherever adequate data quality allows their assessment.

The techniques of causal inference are well satisfied by these results. For comparison, one notes that the E-Value, or Expected value for the tobacco-lung cancer association is 9.0. As noted 21/56 (37.5%) of the E-Values reported herein are above this cut-off and 29/56 (51.87%) exceed 5 which is also a sizeable E-Value.

The E-Value is the value of association required of some extraneous hypothetical unmeasured confounder covariate with both the exposure of concern and the outcome of interest to explain away the observed results. With such high results as are reported in our present paper such a confounder seems most unlikely. Inverse probability weighting is the technique of choice to correctly weight an observational study and turn it into a pseudo-randomized study from which causal inferences can appropriately be drawn. The classic concern with observational studies is that one is in actually comparing “apples with oranges”. The use of inverse probability weighting ensures that everything is “apples” as it were. It is therefore important to appreciate that, although it is true that our study uses several multiple regression techniques, the extensive use of the techniques of quantitative causal inference particularly inverse probability weighting and E-values are the appropriate tools with which to address causal relationships and formally draw causal inferences.

Our study has many strengths including the use of advanced statistical techniques and the formal techniques of causal inference. Its limitations include that we have not had space here to address each disorder separately as those analyses are destined for other manuscripts. In common with all epidemiological studies we do not have individual patient-level data available to us. It is also relevant to observe that the quantitative criteria fulfilled by the analytical procedures in this study are those of causal inference in epidemiology and are widely acknowledged in the discipline. However this is not the same thing as the formal assessment of causality as is done in the controlled experimental laboratory setting, however given that it is not expected to ever be ethical to conduct clinical trials of prenatal exposures these various analytical procedures are the next best thing which can be achieved in clinical populations. Having said that we heartily endorse ongoing research into the many mechanisms of cannabinoid genotoxicity for the same reason that the mechanisms of action of thalidomide continue to be investigated experimentally to better understand its pathophysiology, to remediate its damage, and to develop new lead compounds for novel clinical applications in cancer medicine and elsewhere. As patients can be confused about the impact of early gestational cannabis exposure on developing pregnancies we advocate for the development of a reliable biomarker to quantitate exposure and denominate future studies¹¹.

Some of the statistical techniques used herein also have theoretical limitations. For example it has been noted that inverse probability weighting cannot be used if all the subjects of a certain class at any time during the study must receive a certain exposure condition¹²⁵. Inverse probability weighting has also been observed not to work well with small samples¹²⁶. And its use for dealing with missing data also has methodological weaknesses¹²⁷. However these conditions were not observed in the present dataset and inverse probability was not used to address missing data in these analyses. The interpretation of E-Values is necessarily always subjective and relies on some background knowledge of the subject. For example if an E-Value is reported as five then the judgement must be made as to whether confounding variables are likely to exist which correlate with both the exposure of interest and the outcome of concern of the calculated magnitude to explain away an apparently causal effect. In the present study with median and modal minimum E-Values of 5.65 and 7.75 this seems quite unlikely. Whilst the use of principal components is a common analytical device it can never substitute for detailed investigations of each identified syndrome separately and in detail. For this reason detailed causal modelling and spatiotemporal analyses are indicated on each of the pathologies identified to further investigate the effects reported herein in aggregate.

This report is intended as an introductory overview only and serves the purpose of introducing the subject to readers’ consideration and detailed geospatial and causal inference studies of many congenital anomalies and cancers are indicated to further explore these findings, issues which are indeed the subject of other recent papers^{4,8,10,52,82,83,128–133} and current manuscripts. In conclusion this study of recent US data not only confirms previous findings linking cannabis use with congenital and chromosomal anomalies, but it shows that those impacts are significant at the public health level, likely account for much of their recent rise, explain the worrisome discontinuity and jump from the fourth to the fifth quintiles for cannabis exposure and are consistent both with a rich experimental database and experience from other countries^8,10,37,38. Since findings implicate over 500 MB of the human genome this directly explains the association of cannabis use with many other congenital anomalies and heritable carcinogenesis previously reported. In the context of an exponential dose–response curve for metabolically-genotoxically- and epigenetically-mediated cannabinoid-induced genotoxicity^{14,26,59–61} the rising level of cannabis use induced by cannabis legalization and its severe sequaelae would appear to be more than sufficient contraindication to continued relaxation of the laws surrounding cannabis, risks further compounded by increasingly described heritable neurotoxicity^{128,132,134–139}.