International Journal of Epidemiology, 2015, 566–577
doi: 10.1093/ije/dyv090
Advance Access Publication Date: 7 June 2015
Original article
Mendelian Randomization Causal Analysis
Tobacco smoking is causally associated
with antipsychotic medication use and
schizophrenia, but not with antidepressant
medication use or depression
Marie Kim Wium-Andersen,1,2,3 David Dynnes Ørsted2,3 and
Børge Grønne Nordestgaard1,2,3,4*
1
Department of Clinical Biochemistry, Herlev Hospital, 2Copenhagen General Population Study, Herlev
Hospital, Copenhagen University Hospital, 2730 Herlev, Denmark, 3Faculty of Health and Medical
Sciences, University of Copenhagen, Copenhagen, Denmark and 4Copenhagen City Heart Study,
Frederiksberg Hospital, Copenhagen University Hospital, Frederiksberg, Denmark
*Corresponding author. Department of Clinical Biochemistry, Herlev Hospital, Copenhagen University Hospital, Herlev
Ringvej 75, 2730 Herlev, Denmark. E-mail: [email protected]
Accepted 30 April 2015
Abstract
Background: Tobacco smoking is more common among patients with schizophrenia and
depression than among healthy individuals. We tested the hypothesis that high tobacco
smoking intensity is causally associated with antipsychotic medication use, schizophrenia, antidepressant medication use and/or depression in the general population, and
compared results with those for chronic obstructive pulmonary disease.
Methods: We used self-reported smoking intensity in cigarettes/day and a polymorphism
in the CHRNA3 gene cluster (rs1051730) associated with smoking intensity, on 63 296 20–
100-year-old individuals from the Danish general population; 23 282 were never-smokers
and 40 014 ever-smokers. For schizophrenia, we compared our results with those in the
Psychiatric Genomics Consortium.
Results: In smokers, heterozygotes (CT) and homozygotes (TT) for rs1051730 genotype
had higher smoking intensity compared with non-carriers (CC). Furthermore, in eversmokers homozygotes had increased risk of antipsychotic medication with an odds ratio
(OR) of 1.16 [95% confidence interval (CI) 1.02–1.31] compared with non-carriers,
whereas in never-smokers the corresponding OR was 1.07 (0.87–1.31) (P-interaction:
0.60). Correspondingly, ORs were 1.60 (0.74–3.47) and 1.02 (0.11–9.10) for schizophrenia
(P-interaction: 0.85), 1.02 (0.93–1.13) and 0.99 (0.85–1.15) for antidepressant
medication (P-interaction: 0.87), 0.85 (0.66–1.10) and 1.26 (0.87–1.83) for depression
(P-interaction: 0.30) and 1.31 (1.16–1.47) and 0.89 (0.58–1.36) for chronic obstructive pulmonary disease (P-interaction: 0.16). Odds ratios per rs1051730 allele for schizophrenia
and antipsychotic medication use in ever-smokers in the general population were 1.22
(95% CI: 0.84–1.79) and 1.06 (1.00–1.12). In the Psychiatric Genomics Consortium, the
C The Author 2015; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association
V
566
International Journal of Epidemiology, 2015, Vol. 44, No. 2
567
corresponding OR for schizophrenia was 1.06 (1.04–1.08) in ever- and never-smokers
combined.
Conclusion: Our data suggest that tobacco smoking could influence the development of
psychotic conditions causally, whereas an influence on depression seems unlikely.
Key words: CHRNA3, schizophrenia, Mendelian randomization, general population, genetic variant, rs1051730
Key Messages
• Increased smoking intensity per day and number of pack-years were associated with use of antipsychotic medication,
schizophrenia, use of antidepressant medication and depression in observational analyses.
• The rs1051730 genotype in the nicotinic acetylcholine receptor gene cluster was associated with increased smoking
and with use of antipsychotic medication in ever-smokers, but not in never-smokers. The effect sizes were similar for
schizophrenia.
• The rs1051730 smoking genotype was not associated with use of antidepressant medication or depression in neither
ever- or never-smokers.
• These data suggest that tobacco smoking could influence the development of psychotic conditions causally, whereas
an influence on depression seems unlikely.
Introduction
Tobacco smoking is a well-established causal risk factor of
several somatic diseases including chronic obstructive pulmonary disease, cardiovascular disease and lung cancer.
Smoking has also been associated with major psychiatric
diseases in large observational studies1 showing that tobacco smoking is much more common in patients with
schizophrenia2,3 and depression4,5 than in individuals
without psychiatric disease. A possible explanation is that
patients with psychotic/depressive symptoms increase their
smoking to relieve symptoms6 (reverse causation), but it
has also been argued that smoking per se could contribute
causally to high risk of psychiatric symptoms,7 maybe by
affecting neurotransmitter activity in the brain.8 Thus, the
directionality of these associations remains unclear.
A recent genome-wide association study of more than
150 000 individuals identified more than 100 genetic variants associated with schizophrenia,9 including a variant in
the nicotinic acetylcholine receptor gene cluster CHRNA5A3-B4 which is associated with smoking intensity but not
with smoking initiation.10–13 One potential explanation for
this finding is that the association between the genetic variant and schizophrenia is caused by the effect the genetic
variant has on smoking intensity, that is, if smoking per se
contributes causally to risk of schizophrenia.14
We tested the hypothesis that high tobacco smoking is
causally associated with antipsychotic medication use,
schizophrenia, antidepressant medication use and depression in 63 296 individuals from the Danish general
population, using a Mendelian randomization approach.15
The Mendelian randomization approach enables testing of
causality between an exposure (e.g. tobacco smoking) and
an outcome (e.g. schizophrenia and depression) by the use
of genetic variants. Mendelian randomization exploits the
random allocation of genetic variants during gamete formation, which subsequently results in different phenotypes
(i.e. increased smoking intensity in individuals who choose
to smoke). As such, it can be argued that the principle of a
Mendelian randomization study is analogous to the principle of a randomized controlled trial where participants
are randomized to either intervention or placebo, and the
randomization at baseline ensures equal distribution of
confounders and excludes the possibility of reverse causation. Consequently, if tobacco smoking is in fact causally
related to schizophrenia, we would expect smokers with
genetic variants associated with increased smoking to have
a higher risk of schizophrenia or depression. Conversely, in
never-smokers there is no behavioural difference between
participants with different alleles of the variant and we
would therefore not expect to see a difference in the schizophrenia risk.
In order to fully apply the Mendelian randomization approach, we conducted four different analyses. First, we
tested whether high-intensity tobacco smoking was associated with each of the endpoints in observational analyses.
Second, we tested whether a single nucleotide polymorphism close to the CHRNA3 gene, rs1051730, was associated with self-reported smoking intensity. Third, we tested
568
whether the rs1051730 genotype was directly associated
with each of the endpoints, stratified on smoking status.
We included chronic obstructive pulmonary disease as a
positive control. Finally, for genetic association with the
endpoint schizophrenia, we compared our results with
those from the Psychiatric Genomics Consortium.9
Methods
Participants
We examined 63 296 individuals from two independent
general population studies, the Copenhagen General
Population Study (CGPS) 2003–09 (n ¼ 54 379) and the
Copenhagen City Heart Study (CCHS) 1991–94 and/or
2001–03 examinations (n ¼ 8917).16,17 Participants were
20–100 years old, White, of Danish descent and were randomly selected from the national Danish Civil Registration
System18 to represent the Danish general population.
Participants filled in a questionnaire which was reviewed
together with an investigator on the day of attendance.
Furthermore, all participants had a physical examination
performed and had blood samples drawn for biochemical
measurements and DNA extraction. If a participant appeared in more than one study (n ¼ 329), only data from
the first study were included (Supplementary Figure 1,
available as Supplementary data at IJE online). Based on
an individually unique identification number, we used the
national Danish Civil Registration System to register emigration or death for all participants. Herlev Hospital and a
Danish ethical committee (KF-100.2039/91 and H-KF-01144/01) approved the studies. All participants gave written
informed consent.
Tobacco smoking
Based on questionnaire information, participants were divided into current, former and never-smokers. Current or
former smokers were asked about tobacco type (cigarettes,
cheroots, cigars and/or pipe tobacco), amount smoked per
day and duration of smoking in years. Based on this information, the number of cigarettes or equivalent smoked per
day was calculated. Finally, cumulative cigarette consumption was calculated in pack-years (defined as 20 cigarettes
or equivalents smoked per day for a year) for those participants who reported smoking duration.
Genotyping
DNA was isolated from leukocytes and stored at 45 C.
R
We used the TaqManV
method (Applied Biosystems,
11,19
to genotype rs1051730 close to
Foster City, CA, USA)
the CHRNA3 gene at the Department of Clinical
International Journal of Epidemiology, 2015, Vol. 44, No. 2
Biochemistry, Herlev Hospital. Due to re-runs, the genotyping call rate was 99.9%. The genotype was in Hardy–
Weinberg equilibrium.
Endpoints
From the national Danish Register of Medicinal Product
Statistics, we obtained information about every prescription of antipsychotic or antidepressant medication claimed
by study participants from 1995 through 2012. For antipsychotic medication, we used Anatomical Therapeutic
Chemical (ATC) codes N05A (excluding lithium, N05AN,
primarily used for bipolar disorder) and included all participants with a lifetime purchase of antipsychotic medication. For antidepressant medication, we used ATC codes
N06A and only included participants who at some point in
their life had purchased antidepressant medication for a
period of at least 6 continuous months with an average
daily dose of at least 0.75 of a standard World Health
Organization- (WHO)-defined daily dose.20
Diagnoses of schizophrenia, depression and chronic obstructive pulmonary disease were obtained from the national Danish Patient Registry which has information on
all hospital discharge diagnoses from psychiatric and somatic hospitals since 1977 and from emergency rooms and
outpatient clinics since 1995,21 and from the national
Danish Causes of Death Registry which has information
on causes of death on all individuals in Denmark since
1970, including diagnoses at time of death.22
Schizophrenia was classified according to the International
Classification of Diseases Eighth edition (ICD-8) codes
295.0–9 until 1994, and 10th edition (ICD-10) codes
F20.0–9 from 1994 through 2011. Depression was ICD-8
codes 296.0, 296.2, 298.0 and 300.4, and ICD-10 codes
F32 and F33. Chronic obstructive pulmonary disease was
ICD-8 codes 491 and 492, and ICD-10 codes J41–J44.
Covariates
Covariates for adjustment were chosen because they were
associated with smoking intensity and/or the endpoints
studied. Participants reported on alcohol intake (weekly consumption of drinks; 1 drink 12 g alcohol), weekly physical
activity (0–2 h light; 2–4 h light; > 4 h light activity/2–4 h
vigorous; and > 4 h vigorous), level of education after lower
secondary school (no education; shorter education (less than
3 years); basic vocational training (1–3 years); higher education (3 years); university education), level of income (lowest; middle; highest) and civil status (married; unmarried;
separated; widow/widower). Body mass index (BMI) was
measured weight in kilograms divided by measured height in
metrs squared. Plasma levels of C-reactive protein (CRP)
International Journal of Epidemiology, 2015, Vol. 44, No. 2
569
Endpoints:
Antipsychotic medication use
Tobacco smoking:
cigarettes/day
pack-years
1
Schizophrenia
Antidepressant medication use
Depression
Chronic obstructive pulmonary disease
2
3
4
rs1051730 genotype
1:
2:
3:
4:
Psychiatric Genomics
Consorum
Observational analyses: is tobacco smoking intensity associated with the endpoints? (Figure 2)
Genetic analyses: is rs1051730 genotype associated with increased tobacco smoking? (Figure 3)
Genetic analyses: is rs1051730 genotype associated with the endpoints in ever-smokers and never-smokers? (Figure 4)
Genetic analyses: are the results from a previous GWAS similar to our results for the rs1051730 genotype ? (Figure 5)
Figure 1. Study design. Arrow 1 represents the observational analyses between smoking intensity and each of the endpoints examined by logistic regression models (Figure 2). Arrow 2 represents the association between CHRNA3 rs1051730 genotype and smoking intensity (Figure 3;
Supplementary Table 5, available as Supplementary data at IJE online). Arrow 3 represents the association between CHRNA3 rs1051730 genotype
and each of the endpoints in ever- and never-smokers examined by unadjusted logistic regression (Figure 4). GWAS, genome wide association
study.
were measured with a high-sensitivity assay using
latex-enhanced turbidimetry (Dako, Glostrup, Denmark)
or nephelometry (Dade Behring, Deerfield, IL) at the
Department of Clinical Biochemistry, Herlev University
Hospital. Chronic disease was ascertained by collecting information on diagnoses from the national Danish Patient
Registry, the national Danish Cancer Registry and the national Danish Causes of Death Registry on ischaemic heart
disease, myocardial infarction, stroke, diabetes, hypertension,
cancer, pneumonia, chronic obstructive pulmonary disease,
asthma, deep venous thrombosis and pulmonary embolism.
Statistical analyses
Stata version 12.1 (StataCorp, College Station, TX) was
used for all statistical analyses. To achieve maximal statistical power, data from the CGPS and the CCHS were combined. However, when the two studies were analysed
separately, results were similar, and we adjusted all analyses for study. For all analyses we used two smoking variables including all current and former smokers combined
(¼ ever-smokers, to maximize statistical power), that is,
cigarettes/day and pack-years. However, we also analysed
current and former smokers separately. We conducted five
different analyses (Figure 1).
First, we tested whether smoking intensity as cigarettes/
day (0; 1–10; 11–20; > 20) and number of pack-years (0;
0.1–20; 20.1–40; > 40) were associated with each of the
endpoints, using logistic regression models to calculate
odds ratios (ORs) with 95% confidence intervals (CIs);
these groups each consisted of a large group of participants
and were chosen to achieve sufficient statistical power for
all analyses. We used two different models of adjustment:
(i) age and gender and (ii) multifactorially including age,
gender, alcohol, physical activity, education, income, civil
status, BMI, plasma CRP and chronic disease. For trend
tests, smoking intensity categories were assigned the values
of 1, 2, 3 and 4. We had > 99% complete data on all covariates. Missing values were imputed based on age and gender before multifactorial adjustment.23 We used the Stata
command MI to carry out five imputations of each missing
value which were combined using Rubin’s rules.
Second, we tested whether rs1051730 genotype was
associated with smoking intensity as cigarettes/day and
pack-years using a non-parametric P-trend.
Third, we tested whether rs1051730 genotype was directly associated with all endpoints in ever-smokers and in
never-smokers, using unadjusted logistic regression models
(observed genetic risk). Interaction by smoking status was
tested by introducing a two-factor interaction term in the
model, and subsequently comparing the two models using
a likelihood ratio test.
Finally, we calculated odds ratios per rs1051730 allele
for schizophrenia and antipsychotic medication use in
ever-smokers in the CGPS and the CCHS combined, as
well as for schizophrenia in ever- and never-smokers
570
combined in the publicly available genome-wide association study from the Psychiatric Genomics Consortium
[http://www.broadinstitute.org/mpg/ricopili/].9
Results
Baseline characteristics of the 63 296 participants by smoking intensity as cigarettes/day or equivalent are listed in
Supplementary Table 1 and by pack-years in
Supplementary Table 2 (available as Supplementary data
at IJE online). In total, 23 282 participants (37%) were
never-smokers and 40 014 (63%) were ever-smokers. For
ever-smokers, there was a median tobacco smoking of 13
cigarettes/day (range: 0.5–100) and of 17 pack-years
(range: 0.03–300). In total, 3866 (6%) had purchased antipsychotic medication and 67 (0.1%) had a hospital diagnosis of schizophrenia (of these, 9 had not bought
antipsychotic medication), 7362 (12%) had purchased
antidepressant medication for at least 6 months and 1067
(2%) had a hospital diagnosis of depression (of these, 295
had not bought 6 months of antidepressant medication),
and 3481 (5%) had a hospital diagnosis of chronic obstructive pulmonary disease (Supplementary Table 3, available as Supplementary data at IJE online). Higher smoking
intensity was associated with all potential confounders
(Supplementary Tables 1 and 2) and/or with the endpoints
(Supplementary Table 3 A,B), but the rs1051730 genotype
(Supplementary Tables 4 and 5, available as
Supplementary data at IJE online, Supplementary Table 1)
was not associated with any of the confounders. This
means that an association between the genotype and the
endpoint is largely unconfounded, at least for measured
confounders. Mean follow-up was 6 years (range: 0–20).
Median age at onset was 17 years (interquartile range
15–20) for smoking, 63 years (51–75) for antipsychotic
medication use, 53 years (42–65) for schizophrenia, 56
years (47–68) for antidepressant medication use, 68 years
(55–79) for depression and 68 years (60–75) for chronic
obstructive pulmonary disease (Supplementary Figure 2,
available as Supplementary data at IJE online).
Association between smoking intensity and
endpoints
In observational analyses, smoking intensity as cigarettes/
day was associated with all endpoints (P-trend ¼ 3 104
to 7 10270) (Figure 2). Compared with never-smokers,
participants smoking > 20 cigarettes/day had a multifactorially adjusted OR for antipsychotic medication use of
1.79 (95% CI 1.58–2.02). Corresponding ORs were 6.18
(2.77–13.8) for schizophrenia, 1.92 (1.75–2.10) for antidepressant medication use, 1.55 (1.23–1.94) for depression
International Journal of Epidemiology, 2015, Vol. 44, No. 2
and 9.62 (8.20–11.3) for chronic obstructive pulmonary
disease.
Similarly, increasing number of pack-years was associated with all endpoints (P-trend ¼ 1 103 to
<1 10300) (Figure 2). For antipsychotic medication use,
the multifactorially adjusted OR was 1.53 (95% CI
1.36–1.72) for participants with > 40 pack-years vs neversmokers. Corresponding ORs were 4.13 (1.61–10.6) for
schizophrenia, 1.83 (1.66–2.00) for antidepressant medication use, 1.46 (1.19–1.80) for depression and 12.5 (10.8–
14.6) for chronic obstructive pulmonary disease.
Association between rs1051730 genotype
and smoking intensity
In total, 11% and 44% of participants were homozygotes
(TT) and heterozygotes (CT) for the rs1051730 genotype
(Figure 3 and Supplementary Table 4). For ever-smokers,
homozygotes and heterozygotes smoked 15.6 and 14.5 cigarettes/day compared with 13.6 cigarettes/day for noncarriers (CC) (P-trend ¼ 1 1047) and had smoked 24.3
and 22.3 pack-years vs 20.6 pack-years for non-carriers
(P-trend ¼ 1 1032).
Association between rs1051730 genotype and
endpoints in 40 014 ever-smokers and 23 282
never-smokers
Among ever-smokers, homozygotes (TT) had increased
risk of antipsychotic medication use with an unadjusted
OR of 1.16 (95% CI 1.02–1.31) (P-trend ¼ 0.05 from CC
to CT to TT genotype) compared with non-carriers (CC)
(Figure 3). Correspondingly, ORs were 1.60 (0.74–3.47)
for schizophrenia (P-trend ¼ 0.30), 1.02 (0.93–1.13) for
antidepressant medication use (P-trend ¼ 0.75), 0.85
(0.66–1.10) for depression (P-trend ¼ 0.44) and 1.31
(1.16–1.47) for chronic obstructive pulmonary disease
(P-trend ¼ 3 106). Corresponding analyses stratified
on current and former smokers are shown in
Supplementary Figure 3 (available as Supplementary data
at IJE online).
When we examined never-smokers, rs1051730 genotype was not associated with antipsychotic medication
(P-trend ¼ 0.73), schizophrenia (P-trend ¼ 0.86), antidepressant medication use (P-trend ¼ 0.89), depression
(P-trend ¼ 0.13) or chronic obstructive pulmonary disease
(P-trend ¼ 0.73) (Figure 4). For antipsychotic medication
use, the unadjusted OR was 1.07 (0.87–1.31) for homozygotes compared with non-carriers. Corresponding ORs
were 1.02 (0.11–9.10) for schizophrenia, 0.99 (0.85–1.15)
for antidepressant medication use, 1.26 (0.87–1.83) for depression, and 0.89 (0.58–1.36) for chronic obstructive
International Journal of Epidemiology, 2015, Vol. 44, No. 2
Smoking
intensity
No.
Events
Events /
10,000
participants
571
Multifactorially adjusted
Age and gender adjusted
Antipsychotic medication use
Cigarettes/ day
0
23282
1-10
17931
11-20
16342
>20
5741
pack-years
0
23282
0.1-20
22595
20.1-40
10906
>40
5637
1071
1128
1150
517
460
629
704
901
P-trend: 2*10-40
1.00
1.14
1.51
2.10
(reference)
(1.05-1.25)
(1.38-1.65)
(1.87-2.35)
P-trend: 9*10-22
1.00
1.11
1.33
1.79
(reference)
(1.02-1.21)
(1.21-1.46)
(1.58-2.02)
1071
1240
897
568
460
549
822
1008
P-trend: 2*10-39
1.00
1.13
1.63
1.91
(reference)
(1.05-1.24)
(1.48-1.79)
(1.71-2.14)
P-trend: 1*10-17
1.00
1.12
1.43
1.53
(reference)
(1.03-1.22)
(1.30-1.58)
(1.36-1.72)
0.5
1.0
1.5
2.0
2.5
Schizophrenia
Cigarettes / day
0
23282
1-10
17931
11-20
16342
>20
5741
pack-years
0
23282
0.1-20
22595
20.1-40
10906
>40
5637
3.0
Schizophrenia
0.5
1.0
1.5
2.0
2.5
3.0
10
15
17
25
4
8
10
44
P-trend: 2*10-7
1.00
2.09
2.16
8.61
(reference)
(0.93-4.69)
(0.98-4.75)
(4.00-18.5)
P-trend: 3*10-5
1.00
2.02
1.84
6.18
(reference)
(0.88-4.62)
(0.81-4.17)
(2.77-13.8)
10
20
20
14
4
9
18
25
P-trend: 1*10-6
1.00
1.89
4.59
7.15
(reference)
(0.88-4.06)
(2.09-10.1)
(2.93-17.5)
P-trend: 1*10-3
1.00
1.94
3.42
4.13
(reference)
(0.89-4.23)
(1.49-7.89)
(1.61-10.6)
0
5
10
15
0
20
2
4
6
8 10 12 14 16
Antidepressant medication use
Cigarettes / day
0
23282
1-10
17931
11-20
16342
>20
5741
pack-years
0
23282
0.1-20
22595
20.1-40
10906
>40
5637
2063
2162
2269
868
886
1206
1388
1512
P-trend: 3*10-97
1.00
1.33
1.73
2.26
(reference)
(1.25-1.42)
(1.62-1.85)
(2.07-2.47)
P-trend: 8*10-58
1.00
1.30
1.56
1.92
(reference)
(1.22-1.39)
(1.46-1.66)
(1.75-2.10)
2063
2585
1683
900
886
1144
1543
1597
P-trend: 1*10-105
1.00
1.33
1.93
2.24
(reference)
(1.25-1.41)
(1.80-2.07)
(2.05-2.44)
P-trend: 6*10-59
1.00
1.30
1.72
1.83
(reference)
(1.22-1.38)
(1.60-1.85)
(1.66-2.00)
0.5
1.0
1.5
2.0
2.5
3.0
0.5
1.0
1.5
2.0
2.5
3.0
Depression
Cigarettes / day
0
23282
1-10
17931
11-20
16342
>20
5741
pack-years
0
23282
0.1-20
22595
20.1-40
10906
>40
5637
296
327
316
128
127
182
193
223
P-trend: 8*10-10
1.00
1.15
1.66
2.07
(reference)
(1.06-1.46)
(1.41-1.96)
(1.66-2.58)
P-trend: 3*10-4
1.00
1.13
1.28
1.55
(reference)
(0.96-1.33)
(1.08-1.51)
(1.23-1.94)
296
344
228
166
127
152
209
294
P-trend: 3*10-10
1.00
1.21
1.65
2.20
(reference)
(1.03-1.41)
(1.38-1.97)
(1.80-2.69)
P-trend: 1*10-3
1.00
1.13
1.24
1.46
(reference)
(0.96-1.32)
(1.03-1.48)
(1.19-1.80)
0.5
1.0
1.5
2.0
2.5
0.5
3.0
1.0
1.5
2.0
2.5
3.0
Chronic obstructive pulmonary disease
Cigarettes / day
0
23282
1-10
17931
11-20
16342
>20
5741
pack-years
0
23282
0.1-20
22595
20.1-40
10906
>40
5637
264
938
1552
727
113
523
950
1266
P-trend: <1*10-300
1.00
3.83
9.99
14.2
(reference)
(3.33-4.41)
(8.72-11.4)
(12.2-16.5)
P-trend: 7*10-270
1.00
3.46
7.26
9.62
(reference)
(3.00-3.99)
(6.31-8.35)
(8.20-11.3)
264
802
1172
1151
113
355
1075
2042
P-trend: <1*10-300
1.00
3.14
10.0
19.4
(reference)
(2.73-3.62)
(8.71-11.5)
(16.8-22.4)
P-trend: <1*10-300
1.00
2.89
7.31
12.5
(reference)
(2.50-3.34)
(6.34-8.44)
(10.8-14.6)
0
5
10
15
20
25
Odds ratio (95% confidence interval)
0
2
4
6
8 10 12 14 16
Odds ratio (95% confidence interval)
Figure 2. Observational associations between smoking intensity, antipsychotic medication use, schizophrenia, antidepressant medication use, depression and chronic obstructive pulmonary disease. Based on 63 296 participants from the Copenhagen General Population Study and Copenhagen
City Heart Study combined. Multifactorial adjustment was for age, gender, alcohol consumption, physical activity, education, income, civil status,
body mass index, plasma C-reactive protein and chronic disease.
572
International Journal of Epidemiology, 2015, Vol. 44, No. 2
rs1051730
genotype
Non-carriers (CC)
No. of
participants
Cigarettes/day, mean (SE) ∆ cigarettes/day P trend
18,007 (45%)
0%
Heterozygotes (CT) 17,775 (44%)
7%
Homozygotes (TT)
4,252 (11%)
0
rs1051730
genotype
Non-carriers (CC)
No. of
participants
5
10
15
F
0.4
82
1*10-47
∆ pack-years
P trend
HWE, p(χ2)
R2 (%)
F
1*10-32
0.25
0.3
57
20
Pack-years, mean (SE)
0%
Heterozygotes (CT) 17,380 (44%)
8%
4,167 (11%)
18%
0
0.25
R2 (%)
15%
17,591 (45%)
Homozygotes (TT)
HWE, p(χ2)
5 10 15 20 25 30
Figure 3. rs1051730 genotype and smoking intensity in 40 014 ever-smokers from the Copenhagen General Population Study and Copenhagen City
Heart Study combined. P-values for trend were calculated using Cuzick’s extension of the Wilcoxon rank sum test. HWE P-values are for test of
Hardy–Weinberg disequilibrium. Partial R2 and F values are from the linear regression of logarithmically transformed g/day or pack-years on genotype. HWE, Hardy–Weinberg equilibrium. R2 statistics determine the amount of smoking intensity explained by the genotype, whereas F-statistics
were used to evaluate the strength of the instrument with F > 10 indicating sufficient statistical strength.
pulmonary disease. However, although these results differ
for those among ever-smokers with respect to antipsychotic medication use and chronic obstructive pulmonary disease, there was no formal evidence of statistical
interaction. P for interaction was 0.60 for antipsychotic
medication use, 0.85 for schizophrenia, 0.87 for antidepressant medication use, 0.30 for depression, and 0.16 for
chronic obstructive pulmonary disease.
Comparing results from the Psychiatric Genomics
Consortium and the CGPS and CCHS
Odds ratios per rs1051730 allele for schizophrenia and
antipsychotic medication use in ever-smokers in the CGPS
and the CCHS combined were 1.22 (95% CI: 0.84–1.79)
and 1.06 (1.00–1.12), compared with 1.06 (1.04–1.08) for
schizophrenia in ever- and never-smokers combined in the
Psychiatric Genomics Consortium (Figure 5).
Sensitivity analysis
When we adjusted the genetic analyses for body mass
index, results were similar (data not shown). Furthermore,
there was no interaction of CRP and genotype on the endpoints in ever- and never-smokers (Supplementary Figures
4–5, available as Supplementary data at IJE online).
Finally, when we performed observational analysis prospectively using Cox proportional hazards regression models, results were similar (Supplementary Figure 6, available
as Supplementary data at IJE online).
Discussion
The principal finding of this study of 63 296 participants
(including 40 014 ever-smokers and 23 282 never-smokers)
from the general population is that the rs1051730 genotype in the nicotinic acetylcholine receptor gene cluster
may be associated with higher use of antipsychotic medication in smokers and with schizophrenia overall, but not in
never-smokers; however, there was no statistical evidence
of an interaction between smoking status and genotype
with any of the endpoints. These data suggest that tobacco
smoking could influence the development of psychotic conditions causally, whereas an influence on depression seems
unlikely. These findings are novel.
Mechanistically, the present findings could possibly be
explained by an effect of smoking/nicotine on neurotransmitter activity in the brain;8 however, the precise underlying mechanism is currently unclear. Alternatively, the
association could be mediated by increased inflammation,
as smoking increases inflammation and as inflammation
has been suggested to be a risk factor for schizophrenia.16,24,25 In the present study, the associations between
smoking intensity and all endpoints were still present after
adjusting for plasma C-reactive protein levels (a common
marker of inflammation), rs1051730 genotype was not
associated with plasma C-reactive protein levels and there
was no interaction of C-reactive protein and genotype on
the endpoints. Despite this, we naturally cannot fully exclude inflammation as a mediating link between smoking
and schizophrenia. Another possible explanation for the
International Journal of Epidemiology, 2015, Vol. 44, No. 2
Genotype
∆ smoking
cigarettes/
day
No.
Events
573
Ever-smokers
Antipsychotic medication use
No.
P-trend: 0.05
Events
Never-smokers
P-trend: 0.73
P-interaction: 0.60
CC
0%
18007
1221
1.00 (reference)
10313
477
1.00 (reference)
CT
7%
17755
1238
1.02 (0.94-1.11)
10385
468
0.97 (0.86-1.12)
TT
15 %
4252
336
1.16 (1.02-1.31)
2584
126
1.07 (0.87-1.31)
0.5
Schizophrenia
1.0
1.5
2.0
P-trend: 0.30
0.5
1.0
1.5
2.0
P-trend: 0.86
P-interaction: 0.85
CC
0%
18,007
23
1.00 (reference)
10313
4
1.00 (reference)
CT
7%
17755
25
1.09 (0.62-1.92)
10385
5
1.25 (0.34-4.68)
TT
15 %
4252
9
1.60 (0.74-3.47)
2584
1
1.02 (0.11-9.10)
0.5
Antidepressant medication use
1.0
1.5
2.0
P-trend: 0.75
0.5
1.0
1.5
2.0
P-trend: 0.89
P-interaction: 0.87
CC
0%
18007
2377
1.00 (reference)
10313
917
1.00 (reference)
CT
7%
17755
2346
1.00 (0.94-1.06)
10385
920
1.00 (0.91-1.10)
TT
15 %
4252
576
1.02 (0.93-1.13)
2584
226
0.99 (0.85-1.15)
0.5
Depression
1.0
1.5
2.0
P-trend: 0.44
0.5
1.0
1.5
2.0
P-trend: 0.13
P-interaction: 0.30
CC
0%
18007
347
1.00 (reference)
10313
119
1.00 (reference)
CT
7%
17755
352
1.02 (0.88-1.18)
10385
140
1.18 (0.92-1.51)
TT
15 %
4252
72
0.85 (0.66-1.10)
2584
37
1.26 (0.87-1.83)
0.5
Chronic obstructive pulmonary disease
1.0
1.5
2.0
0.5
1.5
2.0
P-trend: 0.73
P-interaction: 0.16
P-trend: 3*10-6
1.0
CC
0%
18007
1337
1.00 (reference)
10313
118
1.00 (reference)
CT
7%
17755
1471
1.12 (1.04-1.21)
10385
120
1.02 (0.79-1.31)
TT
15 %
4252
409
1.31 (1.16-1.47)
2584
26
0.89 (0.58-1.36)
0.5
1.0
1.5
2.0
Odds ratio (95% confidence interval)
0.5
1.0
1.5
2.0
Odds ratio (95% confidence interval)
Figure 4. Associations between CHRNA3 rs1051730 genotype and antipsychotic medication use, schizophrenia, antidepressant medication use, depression and chronic obstructive pulmonary disease in 40 014 ever-smokers and in 23 282 never-smokers from the Copenhagen General Population
Study and Copenhagen City Heart Study combined. Odds ratios were unadjusted, because genotypes were not associated with measured potential
confounders (see Supplementary Tables 4–6, available at IJE online).
574
International Journal of Epidemiology, 2015, Vol. 44, No. 2
CGPS+CCHS
Schizophenia
(ever-smokers)
Cases / Total no. of participants
Antipsychotic medication
(ever-smokers)
PGC GWAS
Schizophrenia
(ever- and never-smokers)
Odds ratio (95% confidence interval)
57 / 40,014
1.22 (0.84-1.79)
2,795 / 40,014
1.06 (1.00-1.12)
36,989 / 150,064
1.06 (1.04-1.08)
0.8
1.0
1.2
1.4
1.6
1.8
2.0
Odds ratio (95% confidence interval)
Figure 5. Odds ratios for schizophrenia per CHRNA3 rs1051730 allele in 40 014 ever-smokers in the Copenhagen General Population Study (CGPS)
and the Copenhagen City Heart Study (CCHS) combined, and in 150 064 ever- and never-smokers in the Psychiatric Genomics Consortium (PGC) genome-wide association study (GWAS).9 Odds ratios were unadjusted in both studies.
association is that the association between the rs1051730
genotype, smoking and psychosis may be mediated by
increased cannabis use in smokers. Several studies including a systematic review and meta-analysis26 have consistently reported that cannabis increases risk of psychotic
outcomes independently of confounding factors and transient intoxication effects. It is, however, also possible that
tobacco smoking per se could be associated with psychosis
risk through some of the same mechanisms by which
cannabis is believed to cause psychosis, e.g. through modulated activity of dopaminergic, GABAergic and glutamatergic neurons.27
Tobacco smoking has consistently been associated with
schizophrenia2 and depression45 in previous studies, but
whether smoking per se is a causal risk factor in the development of these diseases is unknown. This is contrary to
chronic obstructive pulmonary disease where it is wellknown that smoking is the most important pathogenic factor. The direction of the association between smoking
intensity and depression has been examined in previous
studies: one study including 53 601 participants found no
association between the rs1051730 genotype and depression in smokers.28 Similarly, a study of 6294 pregnant
women found that women carrying the T-allele were less
likely to stop smoking but also less likely to report depressive symptoms.29 Finally, a recent Mendelian randomization meta-analysis of 128 000 participants similarly found
no evidence of a causal association between smoking and
depression.30 These results together with our results do not
support a causal role of tobacco smoking on depression,
even though a causal role has been suggested.7 For schizophrenia, genetic variants in the 15q25 gene cluster have
been associated with schizophrenia,9,31 variation in the
rs1051730 genotype has been associated with negative
symptoms (e.g. low energy, lack of emotional reactivity) in
patients with schizophrenia31,32 and most recently with
schizophrenia per se,9 which raises confidence in the present findings. However, these studies did not include
information on smoking status, which means that whether
this association is likely explained by smoking per se could
not be evaluated.
Mendelian randomization is a variant of a more general method of instrumental variables that has been developed in econometrics. In general, it requires fitting a
specific model in which the genetic factor acts as a proxy
or instrumental variable for the exposure variable (here
smoking). This instrumental variable is then used to predict the outcome (here psychosis/depression). There are,
however, important limitations in this method which relate to the identifiability of the instrumental variable. In
particular, genetic factors which are only weakly related
to the exposure variable may lead to misleading conclusions, that is, if weak instrumental variables are used or if
the variants relate to multiple aspects of the exposure (in
this case if the genetic factor relates to both number of
cigarettes smoked and how heavily they are smoked).
Indeed, because of these potential limitations of an instrumental variable analysis, we on purpose chose not to perform such an instrumental variable analysis, but only
looked at the direct association between genotype and the
outcome.
An important strength of this study is the large sample
size which is required because genetic effects are typically
small. Furthermore, we had information on smoking intensity, rs1051730 genotype and register-based endpoints
of all hospitalizations with schizophrenia and depression,
and all prescriptions of antipsychotic or antidepressant
medication use in a single study. We were also able to
compare results for the psychiatric endpoints with the
positive control of register-based hospitalization with
chronic obstructive pulmonary disease, for which we
know smoking is a causal risk factor. Also, because of the
completeness of the Danish registers we had no losses to
follow-up during a period of up to 20 years. Finally, we
are not aware of other studies which have examined the
association between genetically higher tobacco smoking
International Journal of Epidemiology, 2015, Vol. 44, No. 2
intensity and schizophrenia and use of antipsychotic
medication.
Potential limitations of this study include that we did
not have validated diagnostic scoring scales for schizophrenia or depression. Instead, we used two independent
registers of hospitalizations with schizophrenia or depression and purchase of antipsychotic and antidepressant
medication from Danish pharmacies. All hospital diagnoses of schizophrenia and depression are clinical diagnoses made by physicians based on ICD-8 or ICD-10
criteria, and thereby are likely to have a higher specificity
than medication use. However, using only hospital discharge diagnoses might have underestimated the number
of individuals with schizophrenia or depression, as individuals treated in general practice or by private psychiatrists was not included. As a consequence, we included
information on prescription antipsychotic and antidepressant medication in an attempt to include these participants. Potential limitations for these endpoints are that
both antipsychotics and antidepressants are being prescribed for several conditions other than schizophrenia
and depression, respectively. In Denmark, approved indications for use of antipsychotic medication include treatment of schizophrenia, psychoses, affective disorders with
psychotic symptoms, delirium, organic diseases with
psychotic symptoms, personality disorders, anxiety disorders and certain somatic diseases such as pain disorders,
migraine and cancer. A study of 47 724 patients in the UK
using prescription antipsychotic medication showed that
less than 50% had a diagnosis of a psychotic disorder or
bipolar disorder and other indications for antipsychotic
medication use were anxiety, depression, dementia and
sleep disorders.33 For antidepressants, approved indications include treatment of depression, anxiety disorders,
pain disorders, obsessive compulsive disorder, bulimia
nervosa and smoking cessation. Off-label use might include treatment of sexual disorders, sleeping disorders
and incontinence. A Dutch study of 13 835 patients with
prescription antidepressant medication showed that 46%
received antidepressant medication for depression, and
another 17% for anxiety disorders.34 Accordingly, in an
attempt to exclude participants with symptoms that were
not severe enough to reach the criteria for a diagnosis of
depression or participants treated for conditions other
than depression, we chose only to include participants
who had purchased antidepressants for at least 6 months,
i.e. the recommended duration of continued treatment
after clinical recovery. However, when we included all
participants who had ever received antidepressant medication (20%), results were similar (data not shown).
Another potential limitation of this study is selection
bias, as individuals with psychiatric disease may not
575
participate as often as healthy individuals in a study of the
general population like the present study.16 This can also
be seen in the relatively low point prevalences of schizophrenia and depression (0.1% and 2%, respectively) in our
study, whereas the prevalences in Denmark are 0.5–1%35
for schizophrenia and 16–17% (lifetime prevalence), and
2–3% (point prevalence) for depression.36 In contrast, 6%
of the individuals in our study had purchased antipsychotic
medication. A recent report from the Danish Pharmacy
Association showed that in 2013, 2.3% of the population
was treated with antipsychotic medication, with higher
proportions among older individuals.37 The higher proportion in our study may be because our population mainly
included older individuals.
Another limitation to our endpoints is that the individuals with a diagnosis of schizophrenia or depression in our
study are highly unrepresentative of individuals with
schizophrenia or depression in Denmark in general, as they
may have a milder/less debilitating condition.
Furthermore, possibly due to selection bias the participants
with schizophrenia and depression were quite old at the
time of onset in our study, and studies suggest that participants with late-onset disease may differ from those with
early-onset.38 Furthermore, even though the diagnoses in
the national Danish Patient Registry are of high quality in
general, the validity has not been evaluated for psychiatric
supplementary diagnoses (two-thirds of our diagnoses),
which might have a less certain diagnostic validity. Finally,
for both diseases but especially for schizophrenia, there
can a pronounced diagnostic delay and the diagnoses
might change over time, which also contributes to the misclassification of the diagnoses. However, in general smoking onset preceded onset of all endpoints by many years in
our study, arguing against reverse causation (that is disease
leading to smoking, observationally or in the genetic analyses). We cannot, however, totally exclude that smoking
onset occurred secondary to onset of the endpoints in a
few individuals.
Another possible limitation of this study is the assumption that the rs1051730 genotype should influence the outcomes only through smoking. Studies have suggested that
the rs1051730 genotype is associated with body mass
index in both smokers39 and never-smokers,40 suggesting a
pleiotropic effect of the genotype. In our own studies, however, we have found that the genotype was associated with
body mass index in current smokers only, but not in former or never-smokers.41 We therefore cannot exclude that
the association between antipsychotic medication use,
schizophrenia and the genotype could be mediated through
endocrine pathways (e.g. by lower body mass index) rather
than through smoking, or that the genotype is in linkage
disequilibrium with other variants which may influence
576
risk of antipsychotic medication use, schizophrenia, antidepressant medication use, depression or chronic obstructive pulmonary disease through mechanisms other than
through higher tobacco use. However, when we adjusted
for body mass index in our genetic analyses, results were
similar.
Furthermore, although this study is large, statistical
power may nevertheless be limited given that common
genetic variants account for only a small proportion of
phenotypic variance, and that some of the outcomes
investigated in this study (e.g. schizophrenia) are rare. It
is notable that even for chronic obstructive pulmonary
disease. the P-value for the interaction term is large,
which may be due to limited power. This is probably
because smoking is such a strong risk factor for chronic
obstructive pulmonary disease, so that relatively few
never-smokers will get chronic obstructive pulmonary disease. Thus, to detect an interaction of smoking on the association between genotype and chronic obstructive
pulmonary disorder, we will need more never-smokers
who develop chronic obstructive pulmonary disease.
Also, we did not have information on drug abuse which
could have confounded the observational associations.
However, unless drug abuse is also associated with the
rs1051730 genotype, this might not affect the Mendelian
randomization analysis, but information on cannabis use
in smokers vs non-smokers could help explain whether
the association between the genotypes and psychosis was
mediated by cannabis use. Finally, as we only included
White participants, our results may not necessarily apply
to other races; however, we are not aware of data to suggest that the present results should not be applicable to individuals of most races.
In conclusion, our data suggest that tobacco smoking
could influence the development of psychotic conditions
causally, whereas an influence on depression seems unlikely from the present data.
Supplementary Data
Supplementary data are available at IJE online.
Funding
This work was supported by Herlev Hospital, Copenhagen
University Hospital and the Danish Council for Independent
Research, Medical Sciences (FSS).
Acknowledgements
We thank participants and staff of the Copenhagen General
Population Study and the Copenhagen City Heart Study for their
important contributions.
Conflict of interest: All authors report no conflicts of interest.
International Journal of Epidemiology, 2015, Vol. 44, No. 2
References
1. Lasser K, Boyd JW, Woolhandler S, Himmelstein DU,
McCormick D, Bor DH. Smoking and mental illness: A population-based prevalence study. JAMA 2000;284:2606–10.
2. de Leon J, Diaz FJ. A meta-analysis of worldwide studies demonstrates an association between schizophrenia and tobacco smoking behaviors. Schizophr Res 2005;76:135–57.
3. de Leon J, Diaz FJ. Genetics of schizophrenia and smoking: an
approach to studying their comorbidity based on epidemiological findings. Hum Genet 2012;131:877–901.
4. Breslau N, Kilbey MM, Andreski P. Nicotine dependence and
major depression. New evidence from a prospective investigation. Arch Gen Psychiatry 1993;50:31–35.
5. Glassman AH, Helzer JE, Covey LS et al. Smoking, smoking cessation, and major depression. JAMA 1990;264:1546–49.
6. Spring B, Pingitore R, McChargue DE. Reward value of cigarette
smoking for comparably heavy smoking schizophrenic, depressed, and nonpatient smokers. Am J Psychiatry 2003;160:
316–22.
7. Boden JM, Fergusson DM, Horwood LJ. Cigarette smoking and
depression: tests of causal linkages using a longitudinal birth cohort. Br J Psychiatry 2010;196:440–46.
8. Koranda JL, Cone JJ, McGehee DS, Roitman MF, Beeler JA,
Zhuang X. Nicotinic receptors regulate the dynamic range of
dopamine release in vivo. J Neurophysiol 2014;111:103–11.
9. Schizophrenia Working Group of the Psychiatric Genomics
Consortium. Biological insights from 108 schizophreniaassociated genetic loci. Nature 2014;511:421–27.
10. Berrettini W, Yuan X, Tozzi F et al. Alpha-5/alpha-3 nicotinic receptor subunit alleles increase risk for heavy smoking. Mol
Psychiatry 2008;13:368–73.
11. Kaur-Knudsen D, Bojesen SE, Tybjaerg-Hansen A,
Nordestgaard BG. Nicotinic acetylcholine receptor polymorphism, smoking behavior, and tobacco-related cancer and lung and
cardiovascular diseases: a cohort study. J Clin Oncol 2011;29:
2875–82.
12. Munafo MR, Timofeeva MN, Morris RW et al. Association between genetic variants on chromosome 15q25 locus and objective measures of tobacco exposure. J Natl Cancer Inst 2012;104:
740–48.
13. Thorgeirsson TE, Geller F, Sulem P et al. A variant associated
with nicotine dependence, lung cancer and peripheral arterial
disease. Nature 2008;452:638–42.
14. Flint J, Munafo M. Schizophrenia: genesis of a complex disease.
Nature 2014;511:412–13.
15. Davey Smith G, Ebrahim S. Mendelian randomization:
prospects, potentials, and limitations. Int J Epidemiol 2004;33:
30–42.
16. Wium-Andersen MK, Ørsted DD, Nordestgaard BG. Elevated
C-reactive protein associated with late- and very-late-onset
schizophrenia in the general population: a prospective study.
Schizophr Bull 2014;40:1117–27.
17. Wium-Andersen MK, Ørsted DD, Nielsen SF, Nordestgaard BG.
Elevated C-reactive protein levels, psychological distress, and
depression in 73, 131 individuals. JAMA Psychiatry 2013;70:
176–84.
18. Pedersen CB. The Danish Civil Registration System. Scand J
Public Health 2011;39:22–25.
International Journal of Epidemiology, 2015, Vol. 44, No. 2
19. Kaur-Knudsen D, Nordestgaard BG, Bojesen SE. CHRNA3
genotype, nicotine dependence, lung function and disease in the
general population. Eur Respir J 2012;40:1538–44.
20. World Health Organization Collaborating Center for Drug
Statistics and Methodology. Guidelines for ATC Classification
and DDD Assignment. 15th edn. Oslo: Norwegian Institute of
Public Health, 2012.
21. Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen
JH. The Danish National Hospital Register. A valuable source
of data for modern health sciences. Dan Med Bull 1999;46:
263–68.
22. Helweg-Larsen K. The Danish Register of Causes of Death.
Scand J Public Health 2011;39:26–29.
23. Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med 1991;10:
585–98.
24. Muller N, Myint AM, Schwarz MJ. Inflammation in schizophrenia. Adv Protein Chem Struct Biol 2012;88:49–68.
25. Sommer IE, van Westrhenen R, Begemann MJ, de Witte LD,
Leucht S, Kahn RS. Efficacy of anti-inflammatory agents to improve symptoms in patients with schizophrenia: An update.
Schizophr Bull 2014;40:181–91.
26. Moore TH, Zammit S, Lingford-Hughes A et al. Cannabis use
and risk of psychotic or affective mental health outcomes: a systematic review. Lancet 2007;370:319–28.
27. Pistis M, Ferraro L, Pira L et al. Delta(9)-tetrahydrocannabinol
decreases extracellular GABA and increases extracellular glutamate and dopamine levels in the rat prefrontal cortex: an
in vivo microdialysis study. Brain Res 2002;948:155–58.
28. Bjorngaard JH, Gunnell D, Elvestad MB et al. The causal role of
smoking in anxiety and depression: a Mendelian randomization
analysis of the HUNT study. Psychol Med 2013;43:711–19.
29. Lewis SJ, Araya R, Davey Smith G et al. Smoking is
associated with, but does not cause, depressed mood in pregnancy – a mendelian randomization study. PLoS One 2011;6:
e21689.
30. Taylor AE, Fluharty ME, Bjorngaard JH et al. Investigating the
possible causal association of smoking with depression and anxiety using Mendelian randomisation meta-analysis: the CARTA
consortium. BMJ Open 2014;4:e006141.
577
31. Jackson KJ, Fanous AH, Chen J, Kendler KS, Chen X. Variants
in the 15q25 gene cluster are associated with risk for schizophrenia and bipolar disorder. Psychiatr Genet 2013;23:20–28.
32. Petrovsky N, Quednow BB, Ettinger U et al. Sensorimotor gating
is associated with CHRNA3 polymorphisms in schizophrenia and
healthy volunteers. Neuropsychopharmacology 2010;35:1429–39.
33. Marston L, Nazareth I, Petersen I, Walters K, Osborn DP.
Prescribing of antipsychotics in UK primary care: a cohort study.
BMJ Open 2014;4:e006135.
34. Gardarsdottir H, Heerdink ER, van DL, Egberts AC. Indications
for antidepressant drug prescribing in general practice in the
Netherlands. J Affect Disord 2007;98:109–15.
35. Danish National Board of Health. Sundhedsstyrelsens referenceprogram for skizofreni 2004 [The national board of health’s
recommendations on diagnosis and treatment of schizophrenia
2004]. Copenhagen: Sundhedsstyrelsen, 2004.
36. Danish National Board of Health. Sundhedsstyrelsens referenceprogram for unipolar depression hos voksne 2007 [The national
board of health’s recommendations on diagnosis and treatment
of unipolar depression in adults 2007]. Copenhagen:
Sundhedsstyrelsen, 2007.
37. Danish Pharmacy Association. Borgere i Syddanmark og Region
Sjælland får oftest midler mod psykoser [Citizens in Southern
Denmark and Region Zealand more often receives antipsychotic
medication] Copenhagen: 2015.
38. Wynn Owen PA, Castle DJ. Late-onset schizophrenia: epidemiology, diagnosis, management and outcomes. Drugs Aging
1999;15:81–89.
39. Asvold BO, Bjorngaard JH, Carslake D et al. Causal associations
of tobacco smoking with cardiovascular risk factors: a
Mendelian randomization analysis of the HUNT Study in
Norway. Int J Epidemiol 2014;43:1458–70.
40. Taylor AE, Morris RW, Fluharty ME et al. Stratification by
smoking status reveals an association of CHRNA5-A3-B4 genotype with body mass index in never smokers. PLoS Genet 2014;
10:e1004799.
41. Winsløw UC, Rode L, Nordestgaard BG. High tobacco consumption lowers body weight: a Mendelian randomization study
of the Copenhagen General Population Study. Int J Epidemiol
2015;0:1–11.