Geotemporospatial and causal inferential epidemiological overview and survey of USA cannabis, cannabidiol and cannabinoid genotoxicity expressed in cancer incidence 2003-2017: part 3 - spatiotemporal, multivariable and causal inferential pathfinding and exploratory analyses of prostate and ovarian cancers

Abstract

Background: The epidemiology of cannabinoid-related cancerogenesis has not been studied with cutting edge epidemiological techniques. Building on earlier bivariate papers in this series we aimed to conduct pathfinding studies to address this gap in two tumours of the reproductive tract, prostate and ovarian cancer.

Methods: Age-standardized cancer incidence data for 28 tumour types (including "All (non-skin) Cancer") was sourced from Centres for Disease Control and National Cancer Institute using SEER*Stat software across US states 2001-2017. Drug exposure was sourced from the nationally representative household survey National Survey of Drug Use and Health conducted annually by the Substance Abuse and Mental Health Services Administration 2003-2017 with response rate 74.1%. Federal seizure data provided cannabinoid concentration data. US Census Bureau provided income and ethnicity data. Inverse probability weighted mixed effects, robust and panel regression together with geospatiotemporal regression analyses were conducted in R. E-Values were also calculated.

Results: 19,877 age-standardized cancer rates were returned. Based on these rates and state populations this equated to 51,623,922 cancer cases over an aggregated population 2003-2017 of 124,896,418,350. Inverse probability weighted regressions for prostate and ovarian cancers confirmed causal associations robust to adjustment. Cannabidiol alone was significantly associated with prostate cancer (β-estimate = 1.61, (95%C.I. 0.99, 2.23), P = 3.75 × 10^{- 7}). In a fully adjusted geospatiotemporal model at one spatial and two temporal years lags cannabidiol was significantly independently associated with prostate cancer (β-estimate = 2.08, (1.19, 2.98), P = 5.20 × 10^{- 6}). Cannabidiol alone was positively associated with ovarian cancer incidence in a geospatiotemporal model (β-estimate = 0.36, (0.30, 0.42), P < 2.20 × 10^{- 16}). The cigarette: THC: cannabidiol interaction was significant in a fully adjusted geospatiotemporal model at six years of temporal lag (β-estimate = 1.93, (1.07, 2.78), P = 9.96 × 10^{- 6}). Minimal modelled polynomial E-Values for prostate and ovarian cancer ranged up to 5.59 × 10⁵⁹ and 1.92 × 10¹²⁵. Geotemporospatial modelling of these tumours showed that the cannabidiol-carcinogenesis relationship was supra-linear and highly sigmoidal (P = 1.25 × 10^{- 45} and 12.82 × 10^{- 52} for linear v. polynomial models).

Conclusion: Cannabinoids including THC and cannabidiol are therefore important community carcinogens additive to the effects of tobacco and greatly exceeding those of alcohol. Reproductive tract carcinogenesis necessarily implies genotoxicity and epigenotoxicity of the germ line with transgenerational potential. Pseudoexponential and causal dose-response power functions are demonstrated.

Keywords: Cannabidiol; Cannabigerol; Cannabinoid; Cannabis; Chromosomal toxicity; Congenital anomalies; Dose-response relationship; Epigenotoxicity; Genotoxicity; Mechanisms; Multigenerational genotoxicity; Oncogenesis; Sigmoidal dose-response; Supra-linear dose response; Transgenerational teratogenicity; Δ9-tetrahydrocannabinol.

Fig. 1

Relationship of prostate and ovarian cancer incidence to cannabidiol exposure

Fig. 2

Relationship of prostate and ovarian cancer incidence to cannabidiol exposure by cannabidiol exposure quintile

Fig. 3

Prostate cancer rates by substance exposure

Fig. 4

Prostate cancer rates by estimated cannabinoid exposure

Fig. 5

Map-graph of prostate cancer rates across the USA

Fig. 6

Map-graph of bivariate distribution of prostate cancer and cannabidiol exposure across the USA. Drawn using colorplaner palette

Fig. 7

Geospatial links between various US states (A) edited and (B) Final. These links were used to form the sparse spatial weights matrices used in the geospatial models for prostate and ovarian cancer

Fig. 8

Modelled scaled output values from geospatial models of a comprehensive interactive prostate cancer model lagged to six years

Fig. 9

Relationship of ovarian cancer to various substance exposures

Fig. 10

Relationship of various estimated cannabinoid exposures to ovarian cancer

Fig. 11

Map-graph of ovarian cancer rates across USA over time

Fig. 12

Bivariate map-graph of the relationship between cannabidiol use and the ovarian cancer across USA over time

Fig. 13

Modelled scaled output values from geospatial models of a comprehensive interactive ovarian cancer model lagged to six years