Proportion of patients in south London with first-episode psychosis attributable to use of high potency cannabis: a case-control study

Department of Psychosis Studies (M Di Forti MD, A Marconi MD, E Carra MD, S Fraietta MD, A Trotta MSc, M Bonomo MSc, F Bianconi MSc, P Gardner-Sood PhD, J O’Connor PhD, T R Marques PhD, P Dazzan PhD, Prof A S David MD, F Gaughran MD, Z Atakan MD, C Iyegbe PhD, Prof R M Murray FRS), Department of Health Services and Public Health (S A Stilo MD, Prof C Morgan PhD), Department of Psychological Medicine (V Mondelli PhD, Prof C Pariante PhD), Department of Neuroscience (Prof J Powell DPhil); Department of Addiction (Prof M Lynskey PhD), Institute of Psychiatry, Kings College London, London, UK; and Department of Psychiatry, Mount Sinai School of Medicine, New York, NY, USA (M Russo PhD)


Introduction
Cannabis is the most popular illicit drug in the world. Uruguay was the fi rst country to legalise its use and several US states have done so or are in the process of doing similar. 1 Therefore, any harm caused by cannabis use should be quantifi ed. Prospective epidemiological studies have consistently reported that use of cannabis increases the risk of schizophrenia-like psychosis. 2,3 In the UK, the investigators of the 2012 Schizophrenia Commission 4 concluded that cannabis use is the most preventable risk factor for psychosis, and research that aims to improve estimation of the drug's contribution to illness development should be pursued.
The aspects of exposure to cannabis (eg, age at fi rst use, frequency of use, duration of use) that confer the greatest eff ect on risk of psychosis are unclear. Such information would be valuable for public education and to estimate the proportion of psychosis cases that could be prevented if harmful patterns of cannabis use were removed from the population. The few studies 5, 6 that have tried to estimate the eff ect of cannabis use on the number of new cases of psychosis in specifi c populations have been limited by the scarcity of accurate information on patterns of cannabis use.
The risk of adverse eff ects for mental health and cognition posed by cannabis use has been suggested to depend on the potency of the type of cannabis used. 7 For example, in a previous study 8 of part of the population reported here, we noted that skunk-like types of cannabis, which contain very high concentrations of Δ-9-tetrahydrocannabinol (THC), seemed to have a greater psychotogenic eff ect than did hash (resin), which is known to contain much less THC.
We analysed detailed data for history of cannabis use, aiming to: compare the patterns and types of cannabis used between patients with fi rst-episode psychosis and a population control sample; use the data for pattern of cannabis use to develop a cannabis exposure measure that accurately estimates the risk of psychotic disorders; and calculate the proportion of cases of psychosis in our study area attributable to use of cannabis, particularly high-potency cannabis, if we assumed causality.

Study design and participants
As part of the GAP study, 8 we did a case-control study at the inpatient units of the South London and Maudsley (SLaM) NHS Foundation Trust. We approached all patients aged 18-65 years who presented with fi rst-episode psychosis. We invited patients to participate if they met the International Classifi cation of Diseases 10 criteria for a diagnosis of non-aff ective (F20-F29) or aff ective (F30-F33) psychosis, validated by administration of the Schedules for Clinical Assessment in Neuropsychiatry (SCAN). 9 We excluded individuals who met the criteria for organic psychosis (F09). If patients were too unwell to cooperate, we re-contacted them after the start of treatment.
We recruited controls using internet and newspaper advertisements and by distributing leafl ets at train stations, shops, and job centres. None of the advertising material mentioned cannabis or illicit drug use. Volunteers were administered the Psychosis Screening Questionnaire 10 and were excluded if they met the criteria for a psychotic disorder or if they reported a previous diagnosis of psychotic illness. This study is part of the GAP study, which was granted ethical approval by SLaM and Institute of Psychiatry Local Research Ethics Committee. All case and control individuals included in the study gave written informed consent.

Procedures
We obtained sociodemographic data using the Medical Research Council Schedule. 11 From March, 2006, we took a more detailed history of cannabis use by adding the Cannabis Experience Questionnaire modifi ed version (CEQ mv ) to the assessment. 8,12 From the CEQ mv , we derived information on history of use of tobacco, alcohol, other recreational drugs, and detailed information on cannabis use (age at fi rst use, duration of use, frequency of use, type used).
Measures of cannabis use relevant to the analysis were: lifetime history of cannabis use-ie, had the individual ever used cannabis at any point in their life (no scores 0, yes scores 1); lifetime frequency of cannabis use-ie, the frequency that characterised the individual's most consistent pattern of use (none scores 0, less than once per week every week scores 1, at weekends scores 2, every day scores 3); and type of cannabis used-ie, the type most used by the subject (none scores 0, low potency [hash-type] scores 1, high potency [skunk-type] scores 2). This variable was grouped in accordance with the characteristics of the cannabis samples seized by the Metropolitan Police in London, as reported by Potter and colleagues 13 and the Home Offi ce study (appendix). 14 Finally, we used a seven-item composite cannabis exposure measure derived from the lifetime frequency of use and the most used type (none scores 0, hash less than once per week every week scores 1, hash at weekends scores 2, hash every day scores 3, skunk less than once per week scores 4, skunk at weekends scores 5, skunk every day scores 6) to investigate which patterns of use conferred the greatest risk.

Statistical analysis
We analysed data using Stata 13. We used χ² tests and t tests (or Mann-Whitney U tests) to test for associations between potential confounding variables and between presence of psychotic disorder and exposure to cannabis use. We also used these tests to establish whether missing data for the cannabis use exposure were associated with case-control status and therefore likely to bias the results.
We used logistic regression to analyse whether individual indicators of cannabis use (lifetime use, age at fi rst use, duration and frequency of use, and most used type of cannabis) improved estimation of the likelihood of psychotic disorders (ie, case status), in comparisons of cannabis users with non-users.
We used the punafcc command in Stata 13 to estimate the population attributable fraction (PAF), with confi dence intervals, for each cannabis use variable. The PAF measures the population eff ect of an exposure by providing an estimate of the proportion of disorder that would be prevented if the exposure were removed. However, causality does not have to be proven before the PAF can be estimated, and this causation is not usually established when PAFs are estimated (indeed no single study could ever prove causation). Because the same proportion of disorder attributable to a specifi c risk factor can also be attributable to other factors with which the specifi c risk factor might interact, PAFs for multiple risk factors can add up to more than 100%. Furthermore, the PAF depends on both the prevalence of exposure (ie, measures of cannabis use) in cases and the odds ratio (OR) for the exposure, such that a risk factor with a modest OR can have a major population eff ect if the factor is common.

Role of the funding source
All funders contributed to data collection by providing the salaries of the research workers collecting the data. The funders of the study had no role in study design, data analysis, data interpretation, or writing of the report. All authors had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication.

Results
Between May 1, 2005, and May 31, 2011, we approached 606 patients with fi rst-episode psychosis. Of these 606 patients, 145 (24%) refused to participate. Thus, we recruited 461 patients with fi rst-episode psychosis. Patients who refused to participate were more likely to be men (p<0·004) and of Black Caribbean and Black African ethnic See Online for appendix origin (p=0·001) than were those who consented. Therefore, in all the analyses, we tested for the potential confounding eff ects of ethnic origin and gender. During the same period and from the geographical area served by the clinical units, we recruited 389 control individuals, aged 18-65 years, who were similar to the local population in terms of ethnic origin, education, and employment status (table 1). The later addition of CEQ mv meant that there were data missing on detailed patterns of cannabis use for those participants recruited early in the project. The data we present here are therefore based on 410 (89%) of 461 patients with fi rst-episode psychosis and 370 (95%) of 389 controls for whom we had data for cannabis use.
The patients with fi rst-episode psychosis consisted of more men and were younger than the control group (table 1). As noted previously, 15 patients with fi rst-episode psychosis were also more likely to be of Black ethnic origin (Caribbean or African) compared with controls, and less likely to have completed a high level of education than were controls (table 1).
A larger proportion of patients with fi rst-episode psychosis (184 [45%] of 410 individuals) reported having smoked 100 tobacco cigarettes or more than did controls (60 [16%] of 370 individuals; p<0·0001), but the groups did not diff er in lifetime history of other substance use (p=0·615), or alcohol units consumed per week (p=0·083). Patients with fi rst-episode psychosis were no more likely than were controls to report a lifetime history of ever having used cannabis, but were more likely to use cannabis every day and to mostly use high-potency (skunk-like) cannabis (table 2). A small proportion of cannabis users (3 [0·6%] of 507 individuals) reported having used cannabis more than four days a week and they were included in the every day category.
Among cannabis users, the mean duration of use did not diff er between patients with fi rst-episode psychosis and controls (table 2). On average, both groups started using cannabis in their mid-teens, although distribution of the age at fi rst cannabis use seemed to be skewed (mean 16·1 years, SD 4·2, median 16 years in the patients with fi rst-episode psychosis vs mean 16·6 years, SD 3·2, median 17 years in the control group; Z=2·88; p=0·146). Patients with fi rst-episode psychosis were more likely to start using cannabis at age 15 years or younger than were controls.
When we combined data on frequency of cannabis use and most used type into a single variable, the composite cannabis exposure measure, controls were more likely to be occasional users of low-potency cannabis (hash), and patients with fi rst-episode psychosis were more likely to be daily users of high-potency cannabis (skunk; fi gure 1; p<0·0001).
A logistic regression, adjusted for age, gender, ethnic origin, number of cigarettes smoked, alcohol units and lifetime use of other illicit drugs, education, and employment history, showed that individuals who had ever used cannabis were not at increased risk of psychotic disorder compared with those who had never used   ). Individuals who started using cannabis at ages younger than 15 years had modestly, but signifi cantly, increased risk of psychotic disorders compared with those who never used cannabis (table 3). People who used cannabis or skunk every day were both roughly three times more likely to have a diagnosis of a psychotic disorder than were those who never used cannabis (table 3). We used logistic regression (n=775) to test whether the composite cannabis exposure measure predicted risk of psychotic disorder more accurately than the individual markers, frequency of cannabis use and most used type of cannabis, alone. Individuals who mostly used lowpotency (hash-like) cannabis occasionally (p=0·493), at weekends (p=0·102), or daily (p=0·626) had no increased likelihood of psychotic disorders compared with those who never used cannabis (fi gure 2).
Compared with those who never used cannabis, individuals who mostly used skunk-like cannabis were nearly twice as likely to be diagnosed with a psychotic disorder if they used it less than once per week (p=0·020), almost three times as likely if they used it at weekends (p=0·008), and more than fi ve times as likely if they were daily users (p=0·001; fi gure 2).
Based on the estimated adjusted OR for daily cannabis use (3·04, 95% CI 1·91-7·76), we calculated that, if we assumed causality, 19·3% (13·1-27·0) of psychotic disorders in the study population were attributable to exposure to daily cannabis use. The PAF of psychotic disorders in the study population that were attributable to high potency cannabis use was 24·0% (17·4-30·6) and the PAF for the two exposures combined, skunk use every day, was 16·0% (14·0-20·3; table 4). If causality is assumed, this fi nding suggests that skunk alone was responsible for the largest proportion of new cases (24%) of psychotic disorder in the study population, an eff ect driven by its high prevalence among patients with fi rst-episode psychosis who used cannabis (218 [53%] of 410 patients).

Discussion
The results of our study support our previous conclusions from analysis of part of the sample; 8 use of high-potency cannabis (skunk) confers an increased risk of psychosis compared with traditional low-potency cannabis (hash). Additionally, because of the increased sample size in the present study, we were able to combine information on frequency of use and type of cannabis used into a single measure. This combined measure suggested that the strongest predictor of case-control status (ie, predictor of whether a random individual would be case or control) was daily-skunk use. Figure 2, which shows the adjusted ORs for psychotic disorders for each of the composite cannabis exposure measure groups, shows how the ORs for skunk users increase with the frequency of use.
Samples of skunk seized in the London area in 2005, 13 2008, 14 and more recently, as reported by Freeman and colleagues, 16 contained more THC than did samples of hash, and virtually no cannabidiol. Use of cannabis with a high concentration of THC might have a more detrimental eff ect on mental health than use of a weaker form. Indeed, in line with epidemiological evidence, 2,3 the results of experimental studies 17,18 that investigated the acute eff ects of intravenous administration of THC in non-psychotic volunteers showed that the resulting psychotic symptoms were dependent on the dose. Furthermore, the scarcity of cannabidiol in skunk-like cannabis might also be relevant because evidence suggests that cannabidiol ameliorates the psychotogenic eff ect of THC and might even have antipsychotic properties. 19,20 The presence of cannabidiol might explain our results, which showed that hash users do not have any increase in risk of psychotic disorders compared with non-users, irrespective of their frequency of use. Morgan and colleagues 21 previously reported that, in healthy volunteers who smoked cannabis, individuals with Patients with first-episode psychosis (n=410) Controls (n=370) Odds ratio* (95% CI) p value

Age at fi rst use, years
Never used 1 ·· ≥15 years 0·68 (0·34-1·37) 0·292 <15 years 1·55 (1·00-1·39) 0·048   hair traces of THC and cannabidiol had fewer schizophrenialike symptoms than those with hair traces of THC only. In our results, a combined measure of exposure to cannabis, daily use of high-potency cannabis, predicted a greater risk of psychotic disorders than did the single measures of either frequency or potency. However, a simple yes-or-no question of whether people use skunk might be more useful to identify those at increased risk to develop psychosis because of their cannabis use. In view of the high prevalence of skunk use in our study population, if a causal role for cannabis is assumed, skunk use alone was responsible for 24% of those adults presenting with fi rst-episode psychosis to the psychiatric services in south London.

Frequency of use
South London has one of the highest recorded incidence rates of psychosis in the UK. 22 Boydell and colleagues 23 showed that the incidence of schizophrenia had doubled since 1965, 24 and that one possible contribution to this was the increase in cannabis use among individuals who developed schizophrenia. In the present study, we identifi ed an increased estimate for the PAF accounted for by cannabis (24%) compared with previous studies, which reported PAFs of 6·2% in Germany, 25 8% in New Zealand, 26 and 13·3% in Holland. 5 This fi nding could be caused by, not only the greater use of cannabis, but also the greater use of high-potency (skunk-like) cannabis in south London than in these other countries in earlier periods. 27 Hickman and colleagues 6 suggested that the number of people who need to be treated to stop their cannabis use to prevent one case of schizophrenia is large, but would become substantially lower if more was understood about which individuals are at greatest risk because of their pattern of use or their susceptibility to psychosis. 6 In relation to susceptibility to schizophrenia, Henquet and colleagues 25 calculated that the PAF for individuals in the general population with a predisposition for psychosis at baseline was more than double (14·2%) that of the total population (6·2%). Our data suggest that the potency of the cannabis used also needs to be taken into account in calculations of the PAF.
The strategy we used for control recruitment, based on a variety of advertising strategies rather than on random selection, might have biased the fi ndings. However, the fi nal sample of controls was similar, according to the last UK census data, to the population from which the cases were drawn. Moreover, rather than this approach undersampling individuals who used cannabis, the proportion of controls with a history of cannabis use (63%) was more than the national average (40%) for similar age groups, 28 showing the high prevalence of cannabis use in south London. Furthermore, if we had oversampled individuals who used cannabis, this oversampling would have caused underestimation of the eff ects of cannabis use on risk of psychotic disorders.
A theoretical explanation of why skunk might have been preferred by patients with fi rst-episode psychosis is that, when they began to experience their illness prodrome, these individuals might have sought increased concentrations of THC to self-medicate. However, experimental studies show that THC induces psychotic symptoms, while cannabidiol ameliorates them and reduces anxiety. [16][17][18][19] That people who already have prodromal symptoms would choose a type of cannabis that is high in THC and has little cannabidiol (such as skunk), which might exacerbate their symptoms, rather than a cannabidiol-containing type (such as hash), would seem counterintuitive.
A possible limitation of our study is the absence of data on number of joints or grams used per day. However, because we collected information about use over a period of years and not about present use, the reliability of such detailed information would probably have been confounded by recall bias to a greater extent than was the general description of pattern of use that we obtained. The fact that we were able to collect detailed information on other environmental factors and control for their potential confounding eff ects is a key strength of our study.
Our fi ndings show the importance of raising public awareness of the risk associated with use of high-potency cannabis (panel), especially when such varieties of cannabis are becoming more available. 29 The worldwide