The results of this review show that current use of cannabis (pooled summary OR: 1.62, 95 % CI 1.13–2.31), using cannabis on at least a weekly basis (OR: 1.92, 95 % CI 1.35–2.72) and long duration (>10 years) of cannabis use (OR: 1.50, 95 % CI 1.08–2.09) are all associated with an increased risk of development of TGCT overall, and even more strongly with non-seminoma tumours specifically. There was insufficient evidence to conclude that there is a relationship between seminoma tumours and cannabis use.
Thus, our meta-analyses suggest that a strong association exists between TGCT development and current, chronic and/or frequent cannabis use – particularly non-seminoma development –when compared to those who have never used cannabis.
Biological plausibility of cannabis in testicular carcinogenesis
The primary psychoactive component of the cannabis plant – delta-9-tetrahydrocannabinol, or
THC – stimulates neural cannabinoid receptors, mimicking the action of endogenous cannabanoids (termed
endocannabanoids). The position of these cannabinoid receptors in the basal ganglia, hippocampus, cerebellum and neocortex explains the common neurophysiological effects of cannabis ingestion; however, these receptors are also expressed in peripheral locations, including the testis [
22].
The biological plausibility of the link between cannabis exposure and testicular cancer is thought to be related to disruptions to the hypothalamic–pituitary–testicular axis – an endocrine feedback system which, among other actions, assists with spermatogenesis [
23]. It is thought that cannabis exposure – and subsequent stimulation of cannabinoid receptors – disrupts normal hormone regulation and testicular function, and that this disruption leads to carcinogenesis [
23]. However, evidence regarding the association between regulation of normal testicular function and tumour development remains inconclusive; and given the complex and multifaceted influence of cannabinoid receptor stimulation on biological processes [
9], the path from cannabis exposure to testicular carcinogenesis remains unclear.
Timing of cannabis exposure
The observation of an association between cannabis use and non-seminoma TGCT development – but not seminoma development – is intriguing. As discussed by Skeldon and Goldenberg [
23], this association directs our attention to puberty (rather than later in life) as the key point of exposure. Non-seminoma tumours are typically diagnosed seven [
11] to ten [
12] years earlier than seminoma tumours. Interestingly, one study included in the current review that asked cases and controls about the timing of their first cannabis use showed that those who first-used before the age of 18 years were substantially more likely to develop a non-seminoma TGCT compared to never-users (adjusted OR: 2.80, 95 % CI 1.60–5.10), but that those aged 18 or older were not (OR: 1.30, 95 % CI 0.60–3.20) [
9]. This may suggest that any carcinogenic disruption of interest to the hypothalamic–pituitary–testicular axis occurs during (or before) puberty [
23]; however it is also possible that early initiation of cannabis exposure is a marker of other mediating factors, such as duration and frequency of cannabis use later in life. Another possibility is that since those cases that developed non-seminoma tumours were younger at the time of data collection than those who developed seminoma tumours, they may have been more likely to either recall or report marijuana use. Such a scenario would have the effect of exaggerating the association between cannabis use and non-seminoma development. However, it should be noted that this exaggeration would only occur if the age-matched controls who participated in these studies were not affected by this pattern of differential reporting by age – in other words, if the cannabis use reported by controls was accurate. This is an area that warrants further exploration.
An as-yet unexplored concept regarding the timing of cannabis exposure is the period during prenatal and early childhood development. Best current evidence suggests that TC predisposition is determined prenatally; thus, it is possible that those who positively identify as current, chronic cannabis users are also more likely to have been exposed to cannabis during perinatal and/or early childhood development. In other words, it is possible that primary cannabis use could be a proxy for second-hand exposure to cannabis during the prenatal and/or early childhood period. Such exposure would be congruent with a pre-adulthood disruption to the hypothalamic-pituitary-testicular axis, albeit via a secondary rather than primary source. However this association remains speculative and further research is required regarding the role of non-primary exposure to cannabis during the prenatal and early childhood period as a risk factor for the development of TGCT.
Strengths and weaknesses of included studies
The three case–control studies examined for this review had strengths in a number of areas; however, each of the studies had acknowledged weaknesses, one of these being the ascertainment of cannabis exposure.
For all three studies, exposure to cannabis was measured using self-report – either during a face-to-face interview [
8,
9] or on a written questionnaire [
7]. According to the Newcastle-Ottawa Scale, one of the optimum mechanisms to measure exposure – and ostensibly minimise risk of information bias – is via a structured interview, where the interviewer is blinded to the case/control status of the participant. There is no record in any of the included studies that the interviewers were blinded to the status of the participant. The importance of this is that we do not whether (and to what extent) the association between cannabis exposure and TC was affected by interviewer bias (i.e., the interviewer knowing the case/control status of the participant, and inadvertently leading the participant toward certain answers). However, it would seem unlikely that interviewer bias could explain all or even some of the observed associations between cannabis use and TGCT development; for example, it is difficult to imagine a scenario where knowledge of case/control status would cause interviewers to inadvertently lead those with non-seminoma tumours toward one response, and those with seminoma tumours to another.
In the presence of an association between current cannabis use and testicular cancer development, it would also be desirable to validate self-reported current (or non-current) use via an appropriate specimen-based test. [
24] However the absence of a valid and easily-obtainable biomarker that does not involve the participant providing a urine sample may render such an approach untenable. It is possible that the use of self-report only will underestimate current use of cannabis [
24‐
26]; however there is also some evidence that self-report is an efficacious means of classifying current (or recent) exposure to cannabis among men of similar age to participants of the three included studies [
27].
If the cases and controls are equally likely to either under- or over-report cannabis exposure, then the impact on the observed association between cannabis use and TGCT development would likely be to attenuate it. However, if TGCT cases are more likely to report/recall cannabis use than controls – because of concern that cannabis or similar exposures might be a cause of their cancer, or a similar reason – then this may serve to exaggerate the reported association away from the null. Of course, it is entirely possible that the same exaggeration could occur if cases reported their use accurately, but controls under-reported their use.
The second major weakness for two of the three included studies was low and differential response rates. In one study, the response rate was substantially lower among the controls than the cases [
9]. If the reported cannabis use was different among those controls who responded compared with those who did not, and if the same differential is not present for the cases who responded and cases who did not respond, this will result in biased OR. For example, if the controls who responded had lower rates of cannabis use than non-responding controls, this will lead to an overestimate of the cannabis-TGCT association. Unusually in a second study, the control group had a substantially higher response rate than the case group [
7]. In this study, the controls were friends of the cases, which may explain their willingness to participate in the study. However it is not clear why the response rate among cases was so low. For this latter study, it may be reasonable to assume that cannabis use might have been more similar between cases and controls than if unrelated controls were used. If this is true, we might expect that the ORs in this study would be biased towards the null. Reassuringly, the results of all three studies were reasonably consistent despite the different potential sources of selection bias.
Finally, when considering the role of cannabis in the development of testicular cancer we must also consider the likely pervasiveness of this exposure. For example, it was estimated in the World Drug Report that 12 % of U.S. residents aged 12 or older had used cannabis in 2012 [
2], with 36 % of U.S. college students reported to have used the drug in 2013 [
28]. Given this pervasiveness among young adults, it is likely that ‘ever-use’ will include many individuals with very low exposure to cannabis – meaning that ever-use is unlikely to be a true measure of meaningful cannabis exposure.
It is also worth noting that of all the exposure variables included in our meta-analysis, the greatest heterogeneity between studies was observed for the ever-use variable (I
2 > 50 %). The source of this heterogeneity is obscure and likely to be multifaceted – but could plausibly be due to heterogeneity between study populations in terms of a) pervasiveness of cannabis ever-use and/or b) willingness to report it. For example, fewer controls in the study by Trabert et al. (55 %) [
7] reported ever-use of cannabis compared to the study by Daling et al. (68 %) [
9].