The association between delusional-like experiences, and tobacco, alcohol or cannabis use: a nationwide population-based survey

Subjects were drawn from the Australian National Survey of Mental Health and Wellbeing 2007 (NSMHWB). Details of the methodology have been published elsewhere [22]. In brief, the NSMHWB was a national face-to-face household survey of community residents aged between 16 and 85 years. Sampling was based on random selection from a stratified, multistage area probability sample of private dwellings. Interviews were carried out by trained interviewers from the Australian Bureau of Statistics, a statutory body responsible for conducting such surveys using ethical protocols that include written informed consent. In total, 8,841 individuals participated in the survey that represents an estimated population of 16 015 000 adults. As the response rate for this survey was lower than expected (60%), extensive non-response analyses were undertaken by the Australian Bureau of Statistics in order to assess the reliability of the survey estimates. As a result, adjustments were made to the weighting strategy (full details available elsewhere http://​www.​abs.​gov.​au[23]).

A modified version of the World Mental Health Survey Initiative of the Composite International Diagnostic Interview (WMH-CIDI 3.0) was used to generate lifetime presence of DLEs, and DSM IV based diagnoses of a wide range of common mental health disorders including anxiety disorders, depressive disorders, and alcohol or drug abuse or dependence.

For the assessment of DLEs, we used the items in Section G designed to screen for possible psychosis. Details of the DLEs are given in Table 1. Briefly, the DLE component was composed of three 'screen' items followed by three 'probe' items. Subjects who responded positively to any of the screen items, were administered the 'probe' items. The items covered the following features of psychotic disorders: delusions of control, thought interference and passivity (Question 1 and 1a); delusions of reference and persecution (Question 2 and 2a); and grandiose delusions (Question 3 and 3a). While endorsement of screen items are more common, subsequent endorsement of probe items is more likely to be predictive of true psychotic features [24].

In order to keep a focus on isolated (subclinical) DLE, and in order to allow comparisons with our previous analyses based on this survey [25‐28], individuals who screened positively for schizophrenia (i.e. respondents who reported 'Yes' to the item "Had been told at any time by a psychiatrist that they had schizophrenia") were excluded from the analyses (n = 68), leaving a total of 8 773 subjects for this study.

The WMH-CIDI instrument also includes checklists related to the presence of physical disorders. Details of the methodology have been published elsewhere [26]. Respondents were asked if they ever had any of six physical illnesses: (a) asthma, (b) gout, rheumatism or arthritis, (c) cancers, (d) diabetes or high blood sugar levels, (e) any heart attack, angina or high blood pressure, and (f) stroke or effects of stroke.

Respondents were asked whether they used tobacco regularly ("current user"), and their age of starting use, and current number of cigarettes or other tobacco products used per day. All persons were assessed for alcohol, and cannabis use disorders, based on DSM-IV diagnoses. Details of DSM-IV diagnoses for alcohol use disorders have been published elsewhere [29]. Briefly, in assessing the presence of DSM-IV alcohol use disorders, participants were initially asked whether they had consumed at least 12 alcohol drinks (containing 10 g of ethanol) in any one year in their life-time. Respondents who responded 'yes' were also asked whether they had ever in their lifetime drank on > = 3 days a week and/or have usually consumed > = 3 drinks on the days they were drinking. Only those who answered this question positively were administered the symptom questions that operationalise the DSM-IV alcohol use/dependence disorder diagnostic criteria. A series of 18 questions operationalised the four alcohol use (major role, legal, hazardous use, and social), and seven alcohol dependence (tolerance, withdrawal, cutdown, larger, time spent, give up and continued use) criteria.

Similar to the assessment of alcohol use disorder, the 2007 NSMHWB collected data on cannabis use disorders. In assessing for a cannabis use disorder, subjects were screened for any lifetime use - those who reported lifetime use were then asked a series of five questions for the diagnosis of use disorders, and 10 questions to establish lifetime diagnosis of dependence.

For the main analyses, we examined the association between reporting any DLE (i.e. at least one of the G screen or probe items endorsed), and the tobacco, cannabis and alcohol related variables. For tobacco users, respondents were asked about (a) age at first use, and (b) current number of smokes per day. We divided responses into quartiles (lowest quartile indicative of earlier age of onset, or highest number of cigarettes use respectively). 'No smoking' was used as the reference group for both variables. For cannabis and alcohol, 'age of onset of use disorder' variables were also divided into quartiles where 'no onset' was used as the reference group.

Based on previous research of factors associated with DLE, in Model 1, we adjusted for sex and age of assessment [30, 31]. As previous studies have reported DLEs to be associated with anxiety and depressive disorders [28], and physical disorders [25] we examined a second model adjusting for these potential confounding factors together with other demographic variables including marital status, migrant status, employment status and educational status. As many individuals use more than one substance [15], in Model 2 we also included adjustment for exposure to 'any other' drug or alcohol abuse/dependence (tobacco, alcohol, cannabis, opioid, stimulant and sedatives).

The sample was weighted to adjust for differential probabilities of selection within households, over-sampling of population subgroups and non-response to match census population distribution on a number of geographic and socio-demographic variables [22]. The initial weights were calibrated against known population estimates. Replicate weight variables were developed using the Jack-knife procedure of replication (i.e., the analysis was repeated after one subject was dropped and then the standard error was derived from the distribution of results from all ''minus one'' resamples) [32]. Analyses were performed using Proc Surveylogistic [33] which is designed to analyse complex survey sample using SAS (version 9.2;Cary, NC: SAS Institute). Chi-square tests for linear trend were used to assess dose-response relationships between the exposure variables and DLE.

About one-sixth (17.97%) of the population currently used tobacco every day (Table 3). Compared to those who did not use tobacco, individuals who smoked everyday were significantly more likely to endorse both DLE screen and probe items (Table 3). Individuals who first used tobacco at 15 years or younger were significantly more likely to endorse DLE. With respect to the number of cigarettes per day, individuals with heavy smoking (more than 22 cigarettes per day) were more likely to endorse DLE probe items. However, the pattern of findings in Model 2 did not suggest a clear dose-response relationship.

Table 4 shows the association between cannabis use disorders and DLE. Over 4% of the Australian adult population met criteria for lifetime cannabis use disorder. Those with cannabis dependence disorder were significantly more likely to endorse DLE. Individuals with (a) age of first use 16 years or younger were three times more likely to endorse DLE items. Compared to nonusers, users with the youngest age of onset (16 years or less) had about ten-fold increased odds of endorse DLE probe items.

Table 5 shows the association between alcohol use disorders and DLE. While 18% of the adult population had lifetime alcohol use disorder, the pattern of association between this exposure and DLE was less consistent across the adjusted models and the exposure categories. Concerning age of onset of these two diagnoses versus DLE, those with early onset (17 years or less) were significantly more likely to endorse DLE probe items, however inspection of the data did not suggest a consistent dose-response across the age of onset categories.

Finally, we undertook several post-hoc analyses in order to explore the robustness of the findings. We examined the relationship between the variables of interest when the harmful use and dependence disorders for both cannabis and alcohol were restricted to past year diagnoses (instead of lifetime ever). The pattern of findings remained unchanged (data not shown). We also examined the association between alcohol use disorders versus DLE when demographic variables such as migrant status, employment status and educational status were removed from Model 2. Again, the pattern of findings remained essentially unchanged (data not shown).

We found several significant associations between tobacco-related variables and DLE. For example, compared to non-smokers, DLE endorsement was more common in those (a) those who use tobacco every day (b) those who start smoking earlier before the age of 15 years those who used either 16-22 or 23 or more cigarettes per day. These associations remained significant after controlling for a set of potential confounding factors related to DLE and smoking. The results are consistent with several prospective studies in which tobacco smoking was found to be associated with later psychotic symptoms [18‐20]. The mechanisms underlying this association remains unclear. Mindful that observational studies are vulnerable to residual confounding, the findings raise several research questions suitable for future study. It is possible that the association between tobaccos use and DLE may reflect an increased propensity of young people with psychotic experiences to commence tobacco use. This may be influenced by intrinsic factors (eg., shared genes) that predispose people with psychotic symptoms to initiate and maintenance of smoking behaviours [34]. It is also feasible that nicotine in cigarettes could modulate mesolimbic dopamine neurotransmitter systems [35, 36], which may increase the vulnerability to DLE in susceptible individuals. Animal studies also show that nicotine promotes the release of other neurotransmitters including acetylecholine, endogeneous opioid peptides, GABA, norepinephrine and serotonin [37] which may influence the pathogenesis of psychotic symptoms.

In addition to the significant association between DLE and cannabis use disorders, individuals with onset of use disorders aged 16 years or younger had increased odds of endorsing DLEs. While numerous prospective studies have found an independent effect of cannabis use on the development of psychotic disorders [3, 4, 6, 9], our findings are congruent with recent evidence showing that cannabis use at a younger age was associated with psychotic experiences [17]. One possible explanation is that early exposure to THC (an active ingredients of cannabis) during critical periods of brain maturation has an impact on the development of several neurotransmitter systems which may interfere crucial process of brain development [38]. Recent evidence suggests that Delta 9-THC modulates mediotemporal and ventrostriatal function that may impair verbal learning and provoke acute psychosis [39, 40]. While it is believed that Delta 9-THC may produce permanent changes in the central nervous system [41], recent studies found a clear improvement in outcome in those who stopped use cannabis after the first psychotic episode [42].

In contrast to association between tobacco smoking or cannabis use and DLE, the pattern of association between alcohol use disorders and DLE was less consistent when adjusted for the presence of other substance use disorders. For example, alcohol dependence was significantly associated with endorsement of screen items, but not probe items. The lack of association with the probe items may reflect lack of power, as these items were less frequently endorsed. There was some evidence that early onset of alcohol use and dependence disorders were associated with DLE endorsement (most noticeably for probe items). While those with comorbid schizophrenia and alcohol use disorders have worse clinical outcomes [43], alcohol use has not been identified as a risk factor for psychosis. It is feasible that alcohol use may contribute to an abnormal persistence of DLE (i.e. while DLE endorsement usually decreases over age, those who use alcohol may have persistent DLE). This research question may be suitable for further scrutiny within longitudinal prospective studies.

The study had several limitations. Importantly, the study was cross-sectional and it was not possible to establish the direction of causality between DLEs and the measures of substance use (we do not have information on the age-of-onset of the DLE). Using a prospective birth cohort and a nested sibling-pair study, we have previously reported an association between early cannabis use and later DLE and hallucinations [4]. We hope to explore the association between early alcohol and tobacco use and psychosis-related outcomes in this prospective cohort in future studies. With respect to the outcome measure, there were no items for hallucinations. However, previous studies have shown a strong association between DLE and hallucinations in general population samples [44‐46]. Finally, because the data were obtained from a household survey some population groups such as homeless, people living in nursing home, hostels etc were not surveyed. The study results may not also be generalised in other countries because of potential differences in cultures and socio-economic structures.