The data used in this study were derived from a nationwide survey of 50 087 Swedish men, who were examined for compulsory military training in 1969–1970. Conscription examination was compulsory for all young Swedish men until recently. However, this cohort is the only one that has retained personal identification on matters related to drug use and other behavioral characteristics, thereby enabling record linkage with Sweden’s National Inpatient Register. Over 98% of the men were aged 18–20
years at conscription, and only 2–3% of them were exempted from conscription, mainly due to a severe mental or physical handicap or a congenital disorder. The conscription procedure took 1½
days for each subject to complete. All were given an IQ test and responded to two questionnaires. The first questionnaire concerned social background, upbringing conditions, friendship, relationships, attitudes, and adjustment at school and work. The second concerned use of alcohol, tobacco and other substances. In addition, all the conscripts underwent a medical examination, and also a structured interview with a psychologist. The ones who reported or presented psychiatric problems were referred to a psychiatrist, and any psychiatric disorders found was diagnosed according to the International Classification of Diseases (8th
A total of 607 subjects with depression (ICD-8; 3004) and 30 with psychosis (ICD-8; 295, 296, 297, 298) were identified during conscription, and they were therefore excluded from the analysis. In order to avoid misclassification of outcome, 11 subjects who had diagnoses of both depression and either schizoaffective disorder or schizophrenia during follow-up were also excluded. Due to missing information on cannabis use, we excluded a further 3 614 individuals. The final analytical sample consisted of 45 087 subjects, born 1949–51.
Permission to use the conscription database for research purposes and to perform the relevant record linkages was granted by the Stockholm Regional Ethical Review Board.
Information on use of cannabis and other drugs was obtained from the conscription questionnaire on alcohol, tobacco and substance use. The questions covered whether subjects had ever used drugs, which drugs had been used from a list of alternatives, the first drug used, the drug most commonly used, and how many times a drug had been used.
The main cannabis measure that we employed was reported level of cannabis use as categorized in the response options in the questionnaire: Never, Once, 2–4, 5–10, 11–50, >50. However, due to the small numbers of cases in the sub-groups analyses, we compared outcomes for those having ever used cannabis (thus collapsing all subjects who reported cannabis use into one category) with those who had never used cannabis, and also compared outcomes for those reporting highest level of use (‘>50 times’) with those who had never used cannabis.
Outcomes and follow-up
The Swedish National Inpatient Register, which records all inpatient admissions to hospitals in Sweden, was used to identify admissions for selected diagnoses from 1973 until 2007. The Swedish register recorded approximately 83% of all psychiatric admissions in 1973, 97% in 1974–1983, and 95% in 1984–1986, and has been virtually complete since 1987. Diagnoses were coded according to the Swedish version of the ICD (ICD-8 during 1965–1986, ICD-9 1987–1996, ICD-10 1997–2007), and divided into the following diagnostic groups: 1. Unipolar depression, 2. Bipolar disorder and affective psychosis, 3. Schizoaffective disorder.
ICD 8: Depressive neurosis (300.40)
ICD 9: Depressive disorder not elsewhere classified (311), neurotic depression (300E)
ICD 10: Depressive episode (F32) excluding 323, and recurrent depressive disorder (F33) excluding 333
Bipolar disorder and affective psychosis:
Affective psychosis consisting of involutional melancholia (296.00)
Manic-depression psychosis consisting of manic type (296.10), depressed type (296,20), circular type (296.30), other (296.88) and unspecified (296.99)
Unipolar affective psychosis consisting of manic or hypomanic episodes (296A), melancholia (296B)
Bipolar affective psychosis consisting of manic type (296
C), melancholia type (296D), mixed form (296E), other (296
W) and unspecified (296X)
Manic episode (F30), bipolar affective disorder (F31), severe depressive episode with psychotic symptoms (F323), and recurrent depressive disorder with current psychotic symptoms (F333)
ICD 8: Schizophrenia, type schizoaffective (295.0)
ICD 9: Schizoaffective form (295
ICD 10: Schizoaffective disorders (F25).
Furthermore, we conducted a specific analysis for the diagnoses of mania, which included the following diagnostic codes:
ICD 8: Manic-depressive psychosis (296.10).
ICD 9: Bipolar affective psychosis, manic type (296
ICD 10: Manic episode (F30, F31.0, F31.1, F31.2)
Data were linked to Sweden’s National Cause of Death Register and the Swedish Migration Register. About 1 300 individuals emigrated from Sweden, and 2 620 died during the follow-up period. The date of first emigration and day of death were used as censoring points. First diagnoses of any of the outcomes above were used as the primary end-points. The mean follow-up period from conscription to censoring of data was 32
years, with a range of 1
day to 35
We selected potential confounding variables on the basis of prior research indicating that they are likely to be associated with both cannabis use and affective outcomes. Relevant variables were obtained from the conscription questionnaires and psychological assessments.
a) Diagnosis of personality disorders assessed by a psychiatrist at conscription: any vs. none.
b) IQ score consisted of four main subtests parts (verbal IQ, visuospatial ability, general knowledge and mechanical ability); these four subtests were aggregated to give an overall standardized intelligence score, ranging 1 to 9 (< 74, 74 to 81, 82 to 89, 90 to 95, 96 to 104, 105 to 110, 111 to 118, 119 to 126, > 126). The IQ test has been described elsewhere in detail [22
].We further transformed the standard-nine values into a composite standard-three scale: highest (111 to
126) vs. middle (90 to 110), lowest (< 74 to 89).
c) The variable “Disturbed behavior in childhood” was derived from questions on truancy from school (once a week, once a month, once per term, occasionally), having been in contacts with police and childcare authorities (once, more, never), running away from home (once, more, never), and having been sanctioned in school (once, more, never). These four questions were aggregated to give an overall standardized composite score ranging from 0–9: very low (0 to 1) vs. low (2 to 3), average (4 to 5), high (6 to 7), very high (8 to 9).
d) The variable “Social adjustment” was derived from questions on popularity at school (1
very popular to 5
unpopular), number of close friends (> 5 to none), being in a relationship with a girl (more than a year, several months, one month or less, no). These three questions were aggregated to give an overall standardized score ranges from 0–11: very good (0 to 2) vs. good (3 to 5), low (6 to 8), very low (9 to 11).
e) The variable “Risky use of alcohol” was derived from questions on high consumption of alcohol: none vs. at least one of the following indicators– consumption of at least 250
g 100% alcohol/week; have taken an eye-opener during a hangover; have been apprehended for drunkenness; have reported being drunk often.
f) The variable “Smoking” was based on self-reported information from questionnaires: 1= >20 cigarettes/day; 2
11–20 cigarettes/day; 3
1–10 cigarettes/day; 4
g) “Early adulthood socioeconomic position (Early adulthood SEP)” was based on information from Statistics Sweden on occupation: 1) high/intermediate nonmanual 2) low nonmanual 3) manual skill/unskilled 4) farmers/self-employed/unclassified.
h) “Use of other drugs” (Mebumal, Opium, Preludin, Morphine) was based on information from questionnaires: any vs. none.
i) “Brought up in a city” was based on self-reported information on upbringing: Rural vs. city with less than 50 000 inhabitants, city with more than 50 000 inhabitants, any one of Sweden’s three large metropolitan areas (Stockholm, Gothenburg, and Malmö).
First, the overall depression was defined as first hospital admission for any diagnosis of unipolar depression, bipolar disorder or affective psychosis. Second, specific analyses were performed for unipolar depression, and bipolar disorder and affective psychosis. Due to variation in the ICD coding system over the years, we placed both bipolar disorder and affective psychosis in the same category. Third, an analysis was performed on the association between cannabis use and schizoaffective disorder. Fourth, we performed a specific analysis of the association between cannabis use and manic disorder.
Cox proportional hazard modeling was used to assess the relative risks of developing the outcomes in relation to cannabis exposure. We explored the effect of each individual confounder on the relationship between cannabis use and each outcome. Crude and adjusted hazard ratio (HR), with 95% confidence interval (CI), was computed by level of cannabis use, and also for each potential confounder. We assessed the proportional hazard assumption between cannabis use and each outcome by using a Kaplan-Meier plot. We tested the equality across strata of each individual confounder to explore whether or not to include them in the final model. For the categorical variables we used the log-rank test of equality across strata; in fact, all the variables listed as possible confounders were retained. The quality of the model was tested by running a logistic regression and calculating Hosmer-Lemeshow’s GOF test which had a value was 0.55, showing that our observed results match the expected of the model population. The analyses were performed in SAS 9.1 for Windows.