Year: 2004
Type: article
Abstract: Sir—The report by Smit et al. (2004) is timely, given the growing longitudinal evidence on the linkages between cannabis use and psychosis/psychotic symptoms. These authors point out that there is clear evidence to suggest some form of causal link in which the heavy use of cannabis increases susceptibility to psychosis, with these increases being more likely in those with a predisposition to psychosis. While the review is scholarly and thoughtful in its conclusions, it a serves to highlight two limitations of epidemiological research, as follows. Establishing a causal link requires the demonstration that in the absence of all other factors, exposure to cannabis leads to increases in rates of psychosis. Such conclusions are best achieved by randomized experiments, but for many outcomes in epidemiology such experiments are not possible. This means that evidence about causation has to be based on correlational designs in which associations between cannabis and psychosis are controlled for observed sources of confounding. Subject to the assumption that the observed confounders adequately represent all confounders it becomes possible to draw causal inferences. The difficulty is that this assumption can always be challenged by suggesting the presence of non-observed covariates. Smit et al. acknowledge this point in their comment ‘The confounding hypothesis still leaves room for further debate’ (p. 10). This raises the issue of what steps might be taken to control omitted confounders in future research. Although it is widely believed that epidemiological studies can only control observed sources of confounding, this is not strictly correct. The first approach to addressing this problem is to employ the discordant twin pairs design (Kendler & Prescott 1998; Lynskey et al. 2002; Lynskey et al. 2003). In the simplest form of this design monozygotic (MZ) twin pairs who are discordant for given risk factors (for instance, cannabis use) may be compared on the outcome measure (psychosis). Because the MZ twin pairs share common genes and common environment, these comparisons control for non-observed genes and common environmental factors. An example of this methodology to the cannabis debate is the study by Lynskey et al. (2003), who used the discordant twin design to test the gateway hypothesis. What this study showed was that in twin pairs discordant for cannabis use, the twin using cannabis was more likely to use other illicit drugs. An approach that may be applied to longitudinal data gathered on singletons is to apply the fixed effects model (Hersch & Stratton 1997; Duncan et al. 1998; Fergusson, Horwood & Swain-Campbell 2002). This model uses the changing properties of longitudinal data to take into account the effects of factors which exert fixed effects on the outcome risks. An example that used this approach was the recent study by Fergusson et al. (2002), which used the fixed effects model to examine the effects of cannabis use on a wide range of outcomes in adolescence. That study showed that even when non-observed fixed confounding factors were taken into account, the increasing use of cannabis was associated with a wide range of adverse psycho-social outcomes. These examples show clearly that further and more searching tests of the confounding hypothesis may be possible by using more sophisticated research designs and analysis. The second limitation of the epidemiological method is that the conclusions drawn are interpretations of the patterns of association observed within population-level data. The limitation of this method is that it usually tells us very little about the underlying causal processes that are inferred from the observed associations. This limitation is clearly evident in the review by Smit et al.; while these authors produce clear evidence of a causal link, there is no account of the underlying neurophysiological processes which lead the heavy consumption of cannabis to be transformed into an increased risk of psychosis. It is now clear that cannabis use has complex effects on brain chemistry that is mediated via the cannabinoid receptors (see for example, Ameri 1999; Leweke et al. 1999; Ujike et al. 2002). What is needed to supplement the review by Smit et al. is a parallel review of the neurophysiological evidence examining the possible pathways by which the heavy use of cannabis may lead to psychotic symptoms.
Source: Addiction
Author: DAVID M. FERGUSSON
Language: English
Subfield: Pharmacology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W1482486252