Open Access Research

Investigating the possible causal

association of smoking with depression
and anxiety using Mendelian
randomisation meta-analysis: the CARTA
consortium
Amy E Taylor,1,2 Meg E Fluharty,1,2 Johan H Bjørngaard,3,4
Maiken Elvestad Gabrielsen,5 Frank Skorpen,5 Riccardo E Marioni,6,7,8
Archie Campbell,7 Jorgen Engmann,9 Saira Saeed Mirza,10 Anu Loukola,11
Tiina Laatikainen,12,13,14 Timo Partonen,15 Marika Kaakinen,16,17 Francesca Ducci,18
Alana Cavadino,19 Lise Lotte N Husemoen,20 Tarunveer Singh Ahluwalia,21,22,23
Rikke Kart Jacobsen,20 Tea Skaaby,20 Jeanette Frost Ebstrup,20
Erik Lykke Mortensen,24 Camelia C Minica,25 Jacqueline M Vink,25
Gonneke Willemsen,25 Pedro Marques-Vidal,26 Caroline E Dale,27 Antoinette Amuzu,27
Lucy T Lennon,28 Jari Lahti,29,30 Aarno Palotie,31,32,33 Katri Räikkönen,30
Andrew Wong,34 Lavinia Paternoster,1,35 Angelita Pui-Yee Wong,36,37
L John Horwood,38 Michael Murphy,39 Elaine C Johnstone,40 Martin A Kennedy,41
Zdenka Pausova,42,43 Tomáš Paus,37,44 Yoav Ben-Shlomo,35 Ellen A Nohr,45
Diana Kuh,34 Mika Kivimaki,9 Johan G Eriksson,30,46,47,48,49,50 Richard W Morris,28
Juan P Casas,27,51 Martin Preisig,52 Dorret I Boomsma,25 Allan Linneberg,20,53,54
Chris Power,19 Elina Hyppönen,19,55,56 Juha Veijola,57 Marjo-Riitta Jarvelin,16,17,58,59,60
Tellervo Korhonen,11,12,15 Henning Tiemeier,61 Meena Kumari,9 David J Porteous,6,7
To cite: Taylor AE, Caroline Hayward,62 Pål R Romundstad,3 George Davey Smith,1,35
Fluharty ME, Bjørngaard JH,
et al. Investigating the possible
Marcus R Munafò1,2
causal association of smoking
with depression and anxiety
using Mendelian randomisation
meta-analysis: the CARTA ABSTRACT
consortium. BMJ Open Strengths and limitations of this study
Objectives: To investigate whether associations of
2014;4:e006141. doi:10.1136/
bmjopen-2014-006141
smoking with depression and anxiety are likely to be ▪ This is the largest Mendelian randomisation
causal, using a Mendelian randomisation approach. study of the relationship between smoking and
Design: Mendelian randomisation meta-analyses depression and anxiety conducted to date.
▸ Prepublication history and
additional material is
using a genetic variant (rs16969968/rs1051730) as a ▪ By using a genetic variant associated with
proxy for smoking heaviness, and observational meta- smoking heaviness as a proxy for smoking
available. To view please visit
the journal (http://dx.doi.org/ analyses of the associations of smoking status and heaviness, bias from confounding is minimised
10.1136/bmjopen-2014- smoking heaviness with depression, anxiety and and findings not affected by reverse causality.
006141). psychological distress. ▪ Measurement of depression and anxiety differed
Participants: Current, former and never smokers of across studies so we were unable to use a con-
Received 21 July 2014 European ancestry aged ≥16 years from 25 studies in sistent definition.
Accepted 12 August 2014 the Consortium for Causal Analysis Research in ▪ While results are consistent with no causal asso-
Tobacco and Alcohol (CARTA). ciation between smoking heaviness and depres-
Primary outcome measures: Binary definitions of sion or anxiety, we cannot rule out the
depression, anxiety and psychological distress possibility of a small effect.
assessed by clinical interview, symptom scales or self-
reported recall of clinician diagnosis.
For numbered affiliations see Results: The analytic sample included up to 58 176 odds of anxiety (95% CI 1.54 to 1.90) and 1.69 times
end of article. never smokers, 37 428 former smokers and 32 028 greater odds of psychological distress (95% CI 1.56 to
current smokers (total N=127 632). In observational 1.83) than never smokers. Former smokers also had
Correspondence to analyses, current smokers had 1.85 times greater odds greater odds of depression, anxiety and psychological
Dr Amy Taylor; of depression (95% CI 1.65 to 2.07), 1.71 times greater distress than never smokers. There was evidence for
amy.taylor@bristol.ac.uk

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 1

 Open Access

positive associations of smoking heaviness with depression, anxiety neurotransmitters.23 These systems function to regulate
and psychological distress (ORs per cigarette per day: 1.03 (95% CI the biological and psychological reactions to stressors.
1.02 to 1.04), 1.03 (95% CI 1.02 to 1.04) and 1.02 (95% CI 1.02 to Similarly, human data have demonstrated elevated corti-
1.03) respectively). In Mendelian randomisation analyses, there was sol levels in smokers compared to non-smokers.24
no strong evidence that the minor allele of rs16969968/rs1051730
Constituents of tobacco smoke inhibit the activity of
was associated with depression (OR=1.00, 95% CI 0.95 to 1.05),
monoamine oxidase, enzymes that are involved in the
anxiety (OR=1.02, 95% CI 0.97 to 1.07) or psychological distress
(OR=1.02, 95% CI 0.98 to 1.06) in current smokers. Results were breakdown of monoamines (including dopamine, sero-
similar for former smokers. tonin and norepinephrine); this effect appears to nor-
Conclusions: Findings from Mendelian randomisation analyses do malise following smoking cessation.25 Animal studies
not support a causal role of smoking heaviness in the development of also indicate that both drugs of abuse (including nico-
depression and anxiety. tine) and environmental stressors appear to trigger
changes in midbrain dopaminergic function.26
Consequently, prolonged smoking may act to sensitise
stress response systems, weakening adaptive coping
INTRODUCTION responses and making smokers more susceptible to emo-
Smoking is highly comorbid with both depression and tional distress in response to environmental stressors.
anxiety across many different populations.1–9 Mendelian randomisation methods allow us to investi-
Furthermore, there is evidence to suggest that tobacco gate causal relationships in humans by using genetic var-
control interventions may not be as effective in popula- iants as proxies for exposures of interest. The principle
tions with mental health conditions; for example, recent of Mendelian randomisation relies on the basic (but
trends in the USA suggest that, since 2004, smoking approximate) laws of Mendelian genetics (segregation
rates have declined less rapidly in individuals with and independent assortment). When these principles
anxiety than in the general population.10 Given the pro- hold, genetic variants, at a population level, will not be
found public health burden of both tobacco-related associated with the confounding factors that generally
disease,11 and depression and anxiety,12 understanding distort conventional observational studies.27 28 In add-
this relationship is of great importance. ition, as an outcome measure cannot alter the genotype
Unfortunately, it is difficult to infer causal links that an individual is born with, these analyses should not
between smoking and depression and anxiety from obser- be biased by reverse causality. A genetic variant, single
vational data, due to confounding. There may be factors nucleotide polymorphism (SNP) number rs16969968, in
associated with both smoking and depression and the CHRNA5-CHRNA3-CHRNB4 (CHRNA5-A3-B4) nicoti-
anxiety, such as other substance use (eg, alcohol), socio- nic receptor subunit gene cluster on chromosome 15
economic adversity and education which cannot be fully has demonstrated robust association with smoking heavi-
accounted for in observational studies.13 In addition, ness within smokers.29–32 The rs16969968 variant is func-
even if a causal association does exist, the direction of the tional and leads to an amino acid change (D398N) in
relationship between smoking and depression and the nicotinic receptor α5 subunit protein.33 The minor
anxiety is unclear.14 Prospective studies have provided evi- (risk) allele of this variant is associated with an average
dence that depressive symptoms are associated with increase in smoking amount of one cigarette per day in
increased likelihood of smoking initiation,2 7 15–17 while smokers, and even more strongly associated with
smoking cessation appears to be associated with a short- increases in cotinine (a metabolite of nicotine)
term increase in depressive symptoms during their quit levels.31 34 35 However, given the known role of the
attempt among a subgroup of smokers, and these indivi- variant in altering receptor function,33 it is likely that
duals have poor smoking cessation outcomes.18 This evi- the greater variance explained for cotinine levels is due
dence is consistent with the popular belief that cigarette to this measure better capturing total tobacco exposure,
smoking can reduce anxiety and improve mood, particu- and not because the variant directly affects nicotine
larly among those experiencing anxiety or low mood (the metabolism.31 There is also good evidence that the
self-medication hypothesis). However, there is also a rs16969968 variant, unlike smoking heaviness, does not
growing body of evidence suggesting that smoking may associate with confounding factors that may distort asso-
contribute to the development of these conditions2 7 19– ciations with health outcomes, for example, socio-
21
and that smoking cessation is associated with improve- economic status and education level.36 37
ments in mental health, including depression and The rs16969968 variant (or its proxy rs1051730) has
anxiety, compared to continued smoking.22 been used as an instrument for smoking heaviness in
Plausible biological mechanisms through which consti- Mendelian randomisation studies to demonstrate that
tuents of tobacco smoke may cause depression and smoking causally lowers body mass index38 and that
anxiety have been described. In animal studies, for maternal smoking during pregnancy lowers offspring
example, there is evidence that nicotine administration birth weight39 (see figure 1 for an illustration of the
produces dysregulation in the hypothalamic-pituitary- Mendelian randomisation approach). Using the
adrenal system, which leads to hypersecretion of cortisol rs1051730 variant, two recent studies have applied
and changes in the activity of associated monoamine Mendelian randomisation to examine the causal

2 Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141

 Open Access

Survey of Health and Development (NSHD), the

Netherlands Twin Registry (NTR), Patch 2, the Rotterdam
Study, the Saguenay Youth Study-Parents (SYS-P) and the
Whitehall II study. Further details of these studies are pro-
vided in online supplementary material.

Genotype
Within each study, individuals were genotyped for one of
two SNPs in the CHRNA5-A3-B4 nicotinic receptor subunit
gene cluster, rs16969968 or rs1051730. These SNPs are in
perfect linkage disequilibrium with each other in
Europeans (R2=1.00 in HapMap 3, http://hapmap.ncbi.
nlm.nih.gov/) and therefore represent the same genetic
signal. Where studies had data available for both SNPs, we
used the SNP that was genotyped in the largest number of
Figure 1 Diagram of Mendelian randomisation analysis of
individuals. Details of genotyping methods within each
smoking and depression/anxiety. The genetic variant study are provided in online supplementary material.
rs16969968/rs1051730 is associated with smoking heaviness
but should not be associated with the confounders of the Measures of depression, anxiety and psychological
association between smoking heaviness and depression/ distress
anxiety. In addition, there is no pathway from depression and Depression and anxiety were assessed by clinical inter-
anxiety to the genetic variant (reverse causality). view, symptom scales or self-reported recall of clinician
diagnosis (see table 1). As some of the scales do not dis-
tinguish between symptoms of depression and anxiety,
relationship of smoking with depression and anxiety.37 38
we used the term ‘psychological distress’ to refer a com-
In one large Norwegian population, the rs1051730
posite phenotype.
variant was not associated with depression or anxiety in
To compare measures across studies, we created two
smokers40; in a British cohort, the rs1051730 variant was
case definitions for each of depression, anxiety and psy-
associated with decreased depression during pregnancy
chological distress (see table 2). According to case defin-
in women who smoked prior to pregnancy.41 These find-
ition 1, individuals were classified as depressed or
ings are not consistent with a causal role of smoking in
anxious if they self-reported clinician diagnosis of depres-
increasing depression or anxiety. To test the robustness
sion or anxiety, met clinical criteria for depression
of these findings, we performed a Mendelian randomisa-
(excluding bereavement where known) or anxiety, or
tion meta-analysis combining data from 25 studies
were above previously published cut points for depression
(n=127 632) in the consortium for Causal Analysis
or anxiety on symptom scales. Individuals were classified
Research in Tobacco and Alcohol (CARTA).
as having psychological distress if they met case definition
1 for depression or anxiety, or if they were above a cut
METHODS point on a general scale for psychiatric symptoms. As not
Study populations all scale measures used for assessing mental health have
We used data on individuals aged ≥16 years and of self- published cut-offs for defining cases, we created a second
reported European ancestry from 25 studies from the definition. According to case definition 2, individuals
CARTA consortium: the 1958 Birth Cohort (1958BC), the were classified as depressed or anxious if they were above
Avon Longitudinal Study of Parents and Children the 90th centile for the specific depression or anxiety
(ALSPAC, including both mothers and children), the scales, and psychologically distressed if they were above
British Regional Heart Study (BRHS), the British the 90th centile on either the depression or anxiety scales
Women’s Heart and Health Study (BWHHS), the or above the 90th centile on the general scales of psychi-
Caerphilly Prospective Study (CaPS), the Christchurch atric symptoms. Where both case definitions 1 and 2 were
Health and Development Study (CHDS), Cohorte available within a study, case definition 1 was used. Full
Lausannoise (CoLaus), the English Longitudinal Study of details of the measures and cut points used are provided
Ageing (ELSA), the National FINRISK Study (FINRISK), in online supplementary table S1.
Generation Scotland: the Scottish Family Health Study For the majority of studies (k=17), diagnoses were
(GS:SFHS), Genomics of Overweight Young Adults based on current depression and anxiety (at the time of
(GOYA) females, the Helsinki Birth Cohort Study measurement). Where current diagnoses were not avail-
(HBCS), Health2006, Health2008, the second wave of the able, diagnoses of depression or anxiety in the previous
Nord-Trøndelag health study (HUNT 2), Inter99, the 12 months or lifetime diagnoses were used. For lifetime
Northern Finland Birth Cohorts (NFBC1966 and diagnoses, if information on age at first diagnosis was
NFBC1986), the National Health and Nutrition collected, individuals reporting diagnoses prior to
Examination Survey (NHANES), the MRC National 16 years of age were excluded.

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 3

Open Access

Table 1 Measures of depression, anxiety and psychological distress in the CARTA studies
Study Psychological Distress Depression Anxiety
1958BC CIS-R CIS-R CIS-R
ALSPAC children CIS-R CIS-R CIS-R
ALSPAC mothers EPDS or CCEI EPDS CCEI
BRHS Clinician diagnosis (lifetime)
BWHHS Clinician diagnosis (lifetime)
CaPS GHQ-30 STAI
CHDS CIDI (previous 12 months) CIDI (previous 12 months) CIDI (previous 12 months)
CoLaus DIGS DIGS DIGS
ELSA CES-D or clinician diagnosis of CES-D (8-item) Clinician diagnosis (lifetime)
anxiety
FINRISK Clinician diagnosis (previous
12 months)
Generation GHQ-28 SCIDI/NP (Lifetime diagnosis)
Scotland
GOYA females Clinician diagnosis (since giving Clinician diagnosis (since giving Clinician diagnosis (since giving
birth) birth) birth)
HBCS CES-D or STAI CES-D (20 items) STAI
Health2006 SCL-90-R SCL-90-R SCL-90-R
Health2008 SCL-90-R SCL-90-R SCL-90-R
HUNT HADS HADS HADS
Inter99 SCL-90-R SCL-90-R SCL-90-R
NFBC1966 SCL-25 SCL-25 SCL-25
NFBC1986 YSR YSR YSR
NHANES DIS (lifetime diagnosis)
NSHD GHQ-28
NTR ASR ASR ASR
Patch 2 SCID (lifetime diagnosis)
Rotterdam CES-D or M-CIDI CES-D (20 items) M-CIDI
SYS-P CES-D or DSM instrument CES-D (12 items) 10 question DSM-based
instrument
Whitehall II GHQ-30
All scales measure current depression and anxiety unless otherwise stated. Clinician diagnosis was assessed by self-reported recall in all
studies.
ALSPAC, Avon Longitudinal Study of Parents and Children; ASR, Adult Self Report; BC, Birth Cohort; BRHS, British Regional Heart Study;
BWHHS, British Women’s Heart and Health Study; CARTA, Causal Analysis Research in Tobacco and Alcohol; CCEI, Crown–Crisp
Experiential Index; CES-D, Centre for Epidemiologic Studies Depression; CHDS, Christchurch Health and Development Study; CIDI,
Composite International Diagnostic Interview; CIS-R, Computerised interview schedule-revised; CoLaus, Cohorte Lausannoise; DIGS,
Diagnostic Interview for Genetic Studies; DIS, Diagnostic Interview Schedule; ELSA, English Longitudinal Study of Ageing; EPDS, Edinburgh
Postnatal Depression Scale; GHQ, General Health Questionnaire; GOYA, Genomics of Overweight Young Adults; HADS, Hospital Anxiety
and Depression Scale; HBCS, Helsinki Birth Cohort Study M-CIDI, Munich version of CIDI; NHANES, National Health and Nutrition
Examination Survey; NSHD, National Survey of Health and Development; NTR, the Netherlands Twin Registry; SCID, Structured Clinical
Interview for DSM-III-R diagnosis; SCIDI/NP, Structured Clinical Interview for DSM-IV Axis disorders non-patient edition; SCL, symptoms
checklist; STAI, State Trait Anxiety Inventory; YSR, Youth Self Report.

Symptom scales were also used as continuous mea- assessment for all studies, with the exception of 1958BC,
sures of depression, anxiety and psychological distress CoLaus and HBCS which used smoking status and
(see online supplementary table S2). To compare across depression/anxiety data collected up to 3 years apart
studies, these were converted to z-scores within each (see online supplementary material). Individuals were
study. Most measures of depression, anxiety and psycho- classified as never, former, current or ever (ie, current
logical distress were strongly right skewed. However, and former combined) cigarette smokers. Where infor-
standard transformations (eg, log, square root) did not mation on smoking frequency was available, current
greatly improve distributions in several of the samples. smokers were restricted to individuals smoking at least
Therefore, z-scores were constructed using the untrans- one cigarette per day. Where information on pipe and
formed data in all samples (z-score=(individual score cigar smoking was available, individuals reporting being
−sample mean)/sample SD). current or former smokers of pipes or cigars but not
cigarettes were excluded from all analyses.
Smoking status For studies with adolescent populations (ALSPAC chil-
Smoking status was self-reported (either by question- dren and NFBC1986), analyses were restricted to current
naire or interview) at the same time as mental health daily smokers who reported smoking at least one

4 Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141

 Open Access

Table 2 Case definitions for depression, anxiety and psychological distress

Case definition 1 Case definition 2
Depression Self-report of clinical diagnosis Scoring >90th centile on specific depression
OR scale
Meeting clinical criteria for depression
OR
Scoring above published cut point on specific
depression scale
Anxiety Self-report of clinical diagnosis Scoring >90th centile on specific anxiety
OR scale
Meeting clinical criteria for anxiety disorders
OR
Scoring above published cut point on specific anxiety
scale
Psychological Depression or anxiety as defined above Depression or anxiety as defined above
distress OR OR
Scoring above published cut point on general scale for >90th centile on general scale for psychiatric
psychiatric symptoms symptoms

cigarette per day (current smokers) and individuals who randomisation: that the gene only operates on the
had never tried smoking (never smokers). outcome through its effects on smoking heaviness
Smoking heaviness in current smokers, measured as (ie, no pleiotropy). If rs16969968/rs1051730 only oper-
cigarettes smoked per day, was collected in some studies ates on an outcome measure through smoking heavi-
as a continuous variable and in some studies as a cat- ness, no association should be observed in never
egorical variable. Further details of the smoking mea- smokers. An additive genetic model was assumed, so
sures collected within each study are provided in the ORs represent the difference in odds of the outcome
online supplementary material. per additional copy of the minor (risk) allele. As a sec-
ondary analysis, Mendelian randomisation analyses were
Statistical analysis performed of the association of rs16969968/rs1051730
Analyses were conducted within each contributing study with z-scores of symptoms scales for depression, anxiety
using Stata or R software, following the same analysis and psychological distress using linear regression strati-
plan. Analyses were restricted to individuals with full fied by smoking status. These analyses were adjusted for
data on depression and anxiety outcomes, smoking age and sex and additionally for use of depression or
status and rs16969968/rs1051730 genotype. anxiety medication where available. For studies with a
Sex-adjusted and age-adjusted associations of smoking survey (NHANES) or family-based design (SYS-P),
status (never, former, current, ever) and smoking heavi- appropriate methods were used to adjust SEs (see online
ness with binary measures of depression, anxiety and supplementary material for further information).
psychological distress were assessed using logistic regres- ALSPAC mothers and children were analysed as separate
sion. For the smoking status analysis, never smokers were samples but, as mothers and children were related, sensi-
used as the reference group. The smoking heaviness tivity analyses were performed excluding each one of
analysis was restricted to current daily smokers, and ORs these samples.
represent differences in odds of the outcome measure Results from individual studies were meta-analysed in
per additional cigarette consumed per day. These ana- Stata (V.11) using the ‘metan’ command. Where there
lyses were restricted to studies with continuous measures was evidence of heterogeneity between studies
of cigarettes per day. (I2>50%), both fixed and random effects analyses were
Within each study, genotype frequencies were tested performed. Within the Mendelian randomisation ana-
for deviation from Hardy-Weinberg equilibrium (HWE) lyses, the Cochran Q statistic was used to test for interac-
using a χ2 exact test. Mendelian randomisation analyses tions between genotype and smoking status on the
of the association between rs16969968/rs1051730 and outcome measures.
binary measures of depression, anxiety and psycho- Analyses were also performed stratified by sex because
logical distress were performed using logistic regression there is some evidence from observational studies that
and adjusted for age and sex. Analyses were performed the association between smoking and mental health out-
stratified by smoking status (never, former, current and comes may differ by sex.42 43 Sex differences in the asso-
ever), because the variant only influences smoking ciation between genotype and outcomes measures were
heaviness in individuals who smoke. The analysis in tested for using meta-regression after taking into
never smokers is a test of a key assumption of Mendelian account potential differences by smoking status.

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 5

Open Access

Figure 2 Age-adjusted and

sex-adjusted association of
smoking status with depression,
anxiety and psychological
distress.

RESULTS psychological distress (figure 3). In age-adjusted and sex-

Descriptive statistics adjusted analyses, a one cigarette per day increase in
In total, data on up to 127 632 individuals were available smoking heaviness was associated with a 1.03-fold (95%
for analysis, including 58 176 never smokers, 37 428 CI 1.02 to 1.04, p<0.001) increase in the odds of having
former smokers and 32 028 current smokers. Overall, depression, a 1.03-fold (95% CI 1.02 to 1.04, p<0.001)
45% of the combined study population was male. The increase in the odds of having anxiety and a 1.02-fold
median age within the contributing studies ranged (95% CI 1.02 to 1.03, p<0.001) increase in the odds of
between 16 and 68 years. The mean prevalence of having psychological distress.
depression, anxiety and psychological distress (using As there was evidence of between-study heterogeneity
case definition 1) was 12.5% (range 6.1–37.5%), 10.2% for analyses of both smoking status and smoking heavi-
(range 2.6–19.9%) and 17.4% (range 8.9–27%), respect- ness, random effects meta-analyses are presented.
ively. Descriptive statistics for each of the study popula- However, results from fixed effects meta-analyses were
tions are provided in the online supplementary table S3. similar (data not shown). Individual study estimates for
The minor allele frequency for rs16969968/rs1051730 observational analyses are provided in online supple-
ranged between 0.31 and 0.39 (see online supplemen- mentary figures S1 and S2.
tary table S4). There was no strong evidence for devi-
ation from HWE in any of the studies ( p values all
≥0.06). Mendelian randomisation analysis
There was no clear evidence that rs16969968/rs1051730
was associated with binary measures of depression in
Observational analysis never (OR per minor allele 1.02, 95% CI 0.97 to 1.06,
Levels of depression, anxiety and psychological distress p=0.47), former (OR 1.00, 95% CI 0.95 to 1.05, p=0.99),
differed by smoking status (see figure 2). In age-adjusted current (OR 1.00, 95% CI 0.95 to 1.05, p>0.99) or ever
and sex-adjusted analyses, current smokers had 1.85 (OR 1.01, 95% CI 0.98 to 1.05, p=0.58) smokers (see
times (95% CI 1.65 to 2.07, p<0.001) greater odds of figure 3). Similarly, there was no clear evidence that
depression, 1.71 times (95% CI 1.54 to 1.90, p<0.001) rs16969968/rs1051730 was associated with binary mea-
greater odds of anxiety and 1.69 times (95% CI 1.56 to sures of anxiety in former (OR 1.02, 95% CI 0.97 to
1.83, p<0.001) greater odds of psychological distress 1.08, p=0.44), current (OR 1.02, 95% CI 0.97 to 1.07,
than never smokers. Former smokers had 1.22 times p=0.42) or ever (OR 1.03, 95% CI 0.99 to 1.07, p=0.16)
(95% CI 1.14 to 1.30, p<0.001) greater odds of depres- smokers. However, in never smokers there was some evi-
sion, 1.23 times (95% CI 1.12 to 1.36, p<0.001) greater dence of a positive association between the minor allele
odds of anxiety and 1.17 times (95% CI 1.11 to 1.25, of rs16969968/rs1051730 and anxiety (OR 1.05, 95% CI
p<0.001) greater odds of psychological distress than 1.01 to 1.10, p=0.03). For psychological distress there was
never smokers. a similar pattern, with no strong evidence for an associ-
Among smokers, smoking heaviness was positively asso- ation between rs16969968/rs1051730 in smokers, but
ciated with levels of depression, anxiety and some evidence of a positive association in never smokers

6 Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141

 Open Access

Figure 3 Age-adjusted and sex-adjusted observational and Mendelian randomisation analyses of association of smoking
heaviness with depression, anxiety and psychological distress. Observational analysis performed using random effects
meta-analysis and Mendelian randomisation analysis performed using fixed effects meta-analysis. Observational analysis
restricted to current smokers.

(OR 1.03, 95% CI 1.00 to 1.07, p=0.09). For all out- number of cigarettes smoked per day and depression
comes, there was no clear statistical evidence for a differ- and anxiety, there was no clear evidence for associations
ence in the effect of rs16969968/rs1051730 between between the CHRNA5-A3-B4 variant and these outcomes
never, former and current smokers ( p values for hetero- in smokers. If heavier smoking were to cause
geneity between never, former and current smokers depression or anxiety, we would expect to see an
from Cochran Q test all >0.57). Individual study esti- increased risk of depression or anxiety per copy of the
mates for observational analyses are provided in online minor allele of rs16969968/rs1051730, which increases
supplementary figure S3. smoking heaviness, in current smokers and potentially
Results were similar for continuous measures of symp- also in former smokers, but no difference in risk for
toms of depression, anxiety and psychological distress never smokers. In our meta-analyses, ORs for the effect
(see online supplementary figures S4 and S5). There of rs16969968/rs1051730 in current, former and ever
was no clear evidence for associations of rs16969968/ smokers were all close to the null, with the CIs for these
rs1051730 with continuous outcomes in any of the estimates all overlapping the null. In addition, we found
smoking categories. no evidence to suggest that the variant was differentially
Finally, there was no clear evidence for sex differences associated with depression or anxiety according to
in either observational or Mendelian randomisation smoking status.
analyses for associations between smoking or smoking- Our results are consistent with those of the two previ-
related genotype and depression or anxiety (data avail- ous Mendelian randomisation studies, which did not
able on request). find evidence that smoking increases depression or
anxiety40 or ante-natal depression.41 Both of these
studies were included in this meta-analysis and the
DISCUSSION HUNT study made up more than half of the study
In the largest Mendelian randomisation study on the sample in some analyses. However, exclusion of either of
association of smoking with depression and anxiety con- these samples did not make a substantial difference to
ducted to date, we found no evidence to suggest that effect estimates (see online supplementary figures S6
smoking causes either depression or anxiety. Despite and S7). These findings suggest that previous findings
higher levels of depression, anxiety and psychological linking smoking to higher levels of depression and
distress in current and former smokers compared to anxiety2 9 19 44 may be due to residual confounding, a
never smokers, and a positive association between the shared vulnerability to both mental disorders and

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 7

Open Access

smoking behaviour,45 or reverse causality (eg, if smokers The rs16969968/rs1051730 variant associates with
smoke in an attempt to alleviate the symptoms of depres- smoking heaviness within smokers but is not an instru-
sion or anxiety or depressed smokers find it more diffi- ment for smoking status (ever smoking vs never
cult to quit). Numerous longitudinal studies have smoking).29 Therefore we cannot rule out the possibility
reported that depressive symptoms in childhood and that being a smoker, rather than smoking heaviness
adolescence are associated with increased risk of could influence likelihood of depression or anxiety.
smoking initiation or progression to tobacco depend- However, we do see an observational association
ence.2 17 19 46–48 At the same time, smoking cessation between smoking heaviness and depression and anxiety
appears to be associated with an acute increase in (figure 3), and a dose-dependent relationship between
depressive symptoms among a subgroup of smokers, and an exposure and an outcome strengthens support for a
these individuals have poor smoking cessation out- causal association.52 We might therefore expect to
comes.18 Taken together, this suggests that people observe an association between rs16969968/rs1051730
experiencing depressive symptoms may smoke (or and depression or anxiety if smoking were to cause
relapse to smoking) in an attempt to self-medicate these these conditions. Furthermore, while rs16969968/
symptoms. A Mendelian randomisation study of the asso- rs1051730 is the strongest genetic contributor to
ciation of genetic variants for depression and anxiety smoking behaviour identified to date,53 this variant only
with smoking behaviour would be required to investigate explains a fraction of the estimated 50% of total vari-
the self-medication hypothesis. However, genetic variants ation in smoking behaviour within a population at any
robustly associated with depression and anxiety have yet one time that is due to genetic factors.54 Further signals
to be identified.49 for smoking heaviness have been identified in the same
Some caution should be taken in completely ruling gene cluster,55 in other nicotinic receptor units and in
out an effect of this variant on depression and anxiety other genes, such as those related to nicotine metabol-
within this analysis. The CIs for the associations of ism like CYP2A6.56 Combining these variants in genetic
rs16969968/rs1051730 with depression and anxiety in risk scores for smoking behaviour in Mendelian random-
current smokers overlap the estimates for the per cigar- isation studies will be an important future direction for
ette per day increase in ORs of depression and anxiety, validation of these results.
so we cannot conclusively say that the Mendelian ran- We would not expect to see an effect of rs16969968/
domisation analysis results differ from the observational rs1051730 on depression or anxiety in never smokers,
results (figure 3). This is the most direct comparison because the variant is not associated with smoking heavi-
that we can make with observational estimates in our ness within these individuals. Thus, this group can be
data, given that the minor allele of rs16969968/ used to test potential bias due to pleiotropy (that the
rs1051730 is associated with an average of one cigarette gene affects more than one exposure) in Mendelian ran-
per day increase in smoking heaviness.34 However, this domisation analyses.38 We did observe some evidence
comparison may be problematic because cigarettes per for an association between the variant and anxiety in
day, a self-reported measure of tobacco exposure, does never smokers, a finding previously reported by the
not take into account variation in smoking topography, HUNT study.40 Removal of HUNT from this analysis did
such as the amount of a cigarette an individual smokes not affect the point estimate, suggesting that this associ-
or the depth of inhalation.50 The CHRNA5-A3-B4 variant ation is not driven solely by the data from this study (see
is an instrument for lifetime tobacco exposure within online supplementary figure S6). However, using case
current smokers, and this is not fully captured by cigar- definition 2 (where available) for anxiety in preference
ettes per day. It has been shown that rs16969968/ to case definition 1 slightly attenuated this association in
rs1051730 explains more of the variance in an objective never smokers (see online supplementary figure S8).
measure of tobacco exposure, cotinine (4%), than in While this may be a chance finding, it is possible that
self-reported cigarettes per day (1%).31 35 This appears rs16969968/rs1051730 or a variant in linkage disequilib-
to be why the variant shows a much stronger association rium with this variant, may affect anxiety directly, not
with lung cancer than predicted from the observed asso- through tobacco consumption. There is some suggestion
ciations with self-reported cigarettes per day.31 Therefore from animal studies that nicotinic acetylcholine recep-
if higher levels of smoking did cause depression or tors may play a role in anxiety (eg, mice lacking the α4
anxiety, we might expect the effects of rs16969968/ subunit show increases in anxiety-related behaviour57).
rs1051730 to be considerably larger than those seen However, there is currently little evidence for this associ-
observationally per cigarette per day. For the same ation in humans, and rs16969968/rs1051730 has not
reason, we did not perform instrumental variable ana- been identified in genome-wide association studies of
lysis to estimate the magnitude of the causal effect of depression or anxiety to date.49 58 59
smoking heaviness on depression or anxiety. It has been It is important to note that stratifying by the measured
demonstrated that using cigarettes per day as an inter- exposure variable in Mendelian randomisation studies
mediate variable in Mendelian randomisation analyses can lead to collider bias.60 61 In this specific analysis, if
using rs16969968/rs1051730 can lead to large biases in both the genetic variant and anxiety cause individuals to
causal effect size estimates.51 smoke, then stratifying on smoking could, in theory,

8 Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141

 Open Access

induce an association between the variant and anxiety.13 CONCLUSION

We do not think that collider bias is likely to be a major In conclusion, our Mendelian randomisation analyses
issue in these analyses because rs16969968/rs1051730 do not support a causal role of smoking heaviness
does not appear to be associated with smoking initiation among smokers in the development of depression and
in this sample (see online supplementary figure S9) or in anxiety. While we cannot directly address the question of
previous studies.29 30 There is, however, as reported previ- whether smoking initiation (ie, starting smoking) plays a
ously in a few specific populations36 62 63 some evidence causal role in relation to these outcomes, we expect that
that the minor allele of rs16969968/rs1051730 is asso- if it did we would also see a dose-dependent relationship
ciated with smoking cessation; ORs of being a current between smoking heaviness and depression and anxiety.
compared to a former smoker were 1.08-fold higher per We see such an association in our observational analyses,
copy of the minor allele (95% CI 1.06 to 1.11) in this but no strong evidence for this in our Mendelian ran-
sample. domisation analyses. Future research should focus on
the possible role of depression and anxiety in increasing
Strengths and limitations susceptibility to smoking. As larger genome-wide associ-
The key strength of this study is the large sample size, ation studies of depression and anxiety emerge, it is
using data on over 125 000 individuals from 25 different likely that genetic variants will be identified that can be
populations. Despite this, we did not have the power to utilised in Mendelian randomisation analyses for this
rule out the possibility of a causal effect. A substantial purpose.
increase in sample size would be required to be confident
that what we observe is a true null association in smokers. Author affiliations
1
We hope that our estimates may be combined with those MRC Integrative Epidemiology Unit (IEU) at the University of Bristol, UK
2
of further studies addressing the same question in future UK Centre for Tobacco and Alcohol Studies, School of Experimental
Psychology, University of Bristol, Bristol, UK
meta-analyses, to provide more definitive answers. 3
Department of Public Health, Faculty of Medicine, Norwegian University of
One of the main limitations is the use of broad defini- Science and Technology, Trondheim, Norway
tions of depression and anxiety rather than clinical defi- 4
Forensic Department, Research Centre Bröset St. Olav’s University Hospital
nitions, which were not available in all studies. It is Trondheim, Trondheim, Norway
5
possible that use of more precise phenotypic measures Faculty of Medicine, Department of Laboratory Medicine, Children’s and
Women’s Health, Norwegian University of Science and Technology,
of depression and anxiety based solely on clinical cri- Trondheim, Norway
teria could yield stronger results because non-differential 6
Centre for Cognitive Ageing and Cognitive Epidemiology, University of
misclassification of a binary outcome is likely to attenu- Edinburgh, Edinburgh, UK
ate associations towards the null.2 64 However, we showed 7
Medical Genetics Section, Centre for Genomic and Experimental Medicine,
the expected observational associations between Institute of Genetics and Molecular Medicine, University of Edinburgh,
Edinburgh, UK
smoking and depression, anxiety and psychological dis- 8
Queensland Brain Institute, The University of Queensland, Brisbane, QLD,
tress. In addition, we used two case definitions and per- Australia
formed a sensitivity analysis using case definition 2 in 9
Department of Epidemiology and Public Health, University College London,
preference (where both were available) which produced London, UK
10
similar results (see online supplementary figure S8). Department of Epidemiology, Erasmus Medical Center, Rotterdam, The
Netherlands
Sensitivity analyses performed excluding lifetime defini- 11
University of Helsinki, Hjelt institute, Helsinki, Finland
tions of depression or anxiety also produced similar 12
University of Eastern Finland, Institute of Public Health & Clinical Nutrition,
results (see online supplementary figure S10). Finally, Kuopio, Finland
13
restricting our analyses to those studies with question- Department of Chronic Disease Prevention, National Institute for Health and
naires based on clinical criteria or self-report of doctor Welfare, Helsinki, Finland
14
Hospital District of North Karelia, Joensuu, Finland
diagnosis produced consistent results, although these 15
Department of Mental Health and Substance Abuse Services, National
analyses were underpowered (see online supplementary Institute for Health and Welfare, Helsinki, Finland
figure S11). 16
Institute of Health Sciences, FI-90014 University of Oulu, Finland
17
Although we analysed depression and anxiety separ- Biocenter Oulu, FI-90014 University of Oulu, Finland
18
ately, these conditions are highly comorbid in the South West London and St George’s Mental Health Trust, London, UK
19
Population, Policy and Practice, UCL Institute of Child Health, University
general population65 66 and symptom scale question-
College London, UK
naires are not adequate to distinguish between 20
Research Centre for Prevention and Health, the Capital Region of Denmark,
them.67 68 In addition, the definition of anxiety we used Denmark
21
encompassed general anxiety disorder, panic and Metabolic Genetics Section, Faculty of Health and Medical Sciences, Novo
phobias. It is possible that these conditions have differ- Nordisk Foundation Centre for Basic Metabolic Research, University of
Copenhagen, Copenhagen, Denmark
ent aetiologies.9 Therefore, we cannot make inferences 22
Copenhagen Prospective Studies on Asthma in Childhood, Faculty of Health
about specific anxiety disorders from these results. and Medical Sciences, University of Copenhagen, Copenhagen, Denmark.
Furthermore, the sample encompasses a wide age range, 23
Danish Pediatric Asthma Center, Gentofte Hospital, The Capital Region,
so it is unlikely that this analysis would be able to Copenhagen, Denmark.
24
capture any age-specific effects of smoking on depres- Institute of Public Health and Center for Healthy Aging, University of
Copenhagen, Denmark
sion and anxiety.

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 9

 Open Access

25
Department of Biological Psychology, Netherlands Twin Register, VU Social and Community Medicine in Bristol University. The authors thank the
University, Amsterdam, The Netherlands Health and Social Care Information Centre (HCSIC) for helping us maintain
26
Department of Internal Medicine, Lausanne University Hospital, Lausanne, long term follow-up with the cohort. They also thank all the men who have
Switzerland given their time to be participants in CaPS. HUNT: Nord-Trøndelag Health
27
Faculty of Epidemiology and Population Health, London School of Hygiene Study (The HUNT Study) is a collaboration between HUNT Research Centre
& Tropical Medicine, London, UK (Faculty of Medicine, Norwegian University of Science and Technology NTNU),
28
Department of Primary Care & Population Health, UCL, London, UK Nord-Trøndelag County Council and the Norwegian Institute of Public Health.
29
Institute of Behavioural Sciences, University of Helsinki, Helsinki, Finland NFBC: The authors thank the late Professor Paula Rantakallio (launch of
30
Folkhälsan Research Centre, Helsinki, Finland NFBCs), and Ms Outi Tornwall and Ms Minttu Jussila (DNA biobanking). The
31
Wellcome Trust Sanger Institute, Cambridge, UK authors would like to acknowledge the contribution of the late Academian of
32
The Medical and Population Genomics Program, The Broad Institute of MIT Science Leena Peltonen. NSHD: The authors are very grateful to the members
and Harvard, Cambridge, Massachusetts, USA of this birth cohort for their continuing interest and participation in the study.
33
Institute for Molecular Medicine Finland (FIMM), University of Helsinki,
Collaborators ALSPAC (AET, MEF, MRM, GDS), HUNT ( JHB, PRR, MEG, FS),
Finland
34 Generation Scotland (REM, AC, DJP, CH), Whitehall/ELSA (MeK, JE, MiK),
MRC Unit for Lifelong Health, Ageing at UCL, UK
35 Rotterdam (HT, SSM), FINRISK (AL, TL, TP, TK), NFBC (M-RJ, MaK, JV, FD),
School of Social and Community Medicine, University of Bristol, Bristol, UK
36 1958BC (AC, CP, EH), Health2006/Health2008/Inter99 (LLNH, RKJ, TS, JFE,
Department of Psychology, University of Toronto, Toronto, Canada
37 TSA, AL, ELM), NTR (CCM, GW, DIB, JMV), COLAUS/PsyCoLaus (PM-V, MP),
Rotman Research Institute, Toronto, Canada
38 BWHHS (CED, AA, JPC), BRHS (RWM, LTL), HBCS ( JL, AP, JGE, KR), NSHD
Department of Psychological Medicine, University of Otago, Christchurch,
(AW, DK), Goya females (EAN, LP), CaPS (YB-S), SYS-P (AP-YW, ZP, TP),
New Zealand
39 CHDS (LJH, MAK), Patch 2 (ECJ, MM).
Childhood Cancer Research Group, University of Oxford, Oxford, UK
40
Department of Oncology, University of Oxford, Oxford, UK Contributors MRM and GDS conceived the study. AET and MEF drafted the
41
Department of Pathology, University of Otago, Christchurch, New Zealand analysis protocol and co-ordinated the data analysis. AET conducted the final
42
Departments of Physiology and Nutritional Sciences, University of Toronto, meta-analyses. AET and MRM drafted the initial manuscript. JHB, MEG, FS,
Toronto, Canada REM, AC, JE, SSM, AL, TL, TP, MK, FD, AC, LLNH, TSA, RKJ, TS, JFE, ELM,
43
Hospital for Sick Children, Toronto, Canada CCM, JMV, GW, PMV, CED, AA, LTL, JL, AP, KR, AW, LP, APYW and LJH all
44
Departments of Psychology and Psychiatry, University of Toronto, Toronto, performed analyses within individual studies. MM, ECJ, MAK, ZP, TP, YBS,
Canada EAN, DK, MiK, JGE, RWM, JPC, MP, DIB, AL, CP, EH, JV, MRJ, TK, HT, MeK,
45
Institute for Clinical Research, University of Southern Denmark, Odense, DJP, CH and PRR all substantially contributed to data acquisition in
Denmark contributing studies. All authors commented on a draft of the manuscript and
46
Department of Medical Genetics, University of Helsinki and University approved the final version.
Central Hospital, Helsinki, Finland
47 Funding 1958BC: Statistical analyses were funded by the Academy of Finland
National Institute for Health and Welfare, Finland
48 (Project 24300796 and SALVE/PREVMEDSYN). DNA collection was funded by
Department of General Practice and Primary health Care, University of
MRC grant G0000934 and cell-line creation by Wellcome Trust grant 068545/
Helsinki, Finland
49 Z/02. This research used resources provided by the Type 1 Diabetes Genetics
Unit of General Practice, Helsinki University Central Hospital, Helsinki,
Consortium, a collaborative clinical study sponsored by the National Institute
Finland
50 of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institute of
Vasa Central Hospital, Vasa, Finland
51 Allergy and Infectious Diseases, National Human Genome Research Institute,
Institute of Cardiovascular Science, University College London, UK
52 National Institute of Child Health and Human Development, and Juvenile
Department of Psychiatry, Lausanne University Hospital, Prilly, Switzerland
53 Diabetes Research Foundation International ( JDRF) and supported by U01
Department of Clinical Experimental Research, Glostrup University Hospital,
DK062418. Funding for the project was provided by the Wellcome Trust under
Denmark
54 the award 076113. Great Ormond Street Hospital/University College London,
Faculty of Health and Medical Sciences, Department of Clinical Medicine,
Institute of Child Health receives a proportion of funding from the Department
University of Copenhagen, Denmark
55 of Health’s National Institute for Health Research (NIHR) (’Biomedical
School of Population Health and Sansom Institute, University of South
Research Centres’ funding). ALSPAC: The UK Medical Research Council and
Australia, Adelaide, Australia
56 the Wellcome Trust (Grant ref: 092731) and the University of Bristol provide
South Australian Health and Medical Research Institute, Adelaide, Australia
57 core support for ALSPAC. This work was supported by the Wellcome Trust
Department of Psychiatry, Oulu University Hospital, Oulu, Finland
58 (grant number 086684) and the Medical Research Council (grant numbers
Department of Epidemiology and Biostatistics, MRC Health Protection
MR/J01351X/1, G0800612, G0802736, MC_UU_12013/1, MC_UU_12013/6).
Agency (HPA) Centre for Environment and Health, School of Public Health,
BRHS: The collection and management of data over the last 34 years of the
Imperial College London, UK
59 BRHS has been made possible through grant funding from UK government
Unit of Primary Care, Oulu University Hospital, Oulu, Finland
60 agencies and charities. BWHHS: The British Women’s Heart and Health Study
Department of Children and Young People and Families, National Institute
has been supported by funding from the British Heart Foundation (BHF)
for Health and Welfare, Oulu, Finland
61 (grant PG/09/022) and the UK Department of Health Policy Research
Department of Epidemiology and Psychiatry, Erasmus Medical Center,
Programme (England) (grant 0090049). The BWHHS HumanCVD data were
Rotterdam, The Netherlands
62 funded by the BHF (PG/07/131/24254). CHDS: The Christchurch Health and
Medical Research Council Human Genetics Unit, Institute of Genetics and
Development Study has been supported by funding from the Health Research
Molecular Medicine, University of Edinburgh, Edinburgh, UK
Council of New Zealand, the National Child Health Research Foundation (Cure
Kids), the Canterbury Medical Research Foundation, the New Zealand Lottery
Acknowledgements ALSPAC: The authors are extremely grateful to all the Grants Board, the University of Otago, the Carney Centre for
families who took part in this study, the midwives for their help in recruiting Pharmacogenomics, the James Hume Bequest Fund, US NIH grant
them, and the whole ALSPAC team, which includes interviewers, computer MH077874, and NIDA grant ‘‘A developmental model of gene-environment
and laboratory technicians, clerical workers, research scientists, volunteers, interplay in SUDs’’ (R01DA024413) 2007–2012. Colaus/PsyCoLaus: The
managers, receptionists and nurses. BRHS: The British Regional Heart Study CoLaus/PsyCoLaus study was supported by four grants of the Swiss National
is a British Heart Foundation (BHF) Research Group. BWHHS: The authors Science Foundation (#105993, 118308, 139468 and 122661), two
thank all BWHHS participants, the general practitioners and their staff who unrestricted grants from GlaxoSmithKline as well as by the Faculty of Biology
have supported data collection since the study inception. CaPS: The Caerphilly and Medicine of the Lausanne University. ELSA: ELSA is funded by the
Prospective Study was conducted by the former MRC Epidemiology Unit National Institute on Aging in the US (R01 AG017644;R01AG1764406S1) and
(South Wales). The Caerphilly archive is now maintained by the School of by a consortium of UK Government departments (including: Department for

10 Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141

 Open Access

Communities and Local Government, Department for Transport, Department National Institutes of Health, HT: Dutch Medical Research Council, TP:
for Work and Pensions, Department of Health, HM Revenue and Customs and Canadian Institutes of Health Research, MK: Medical Research Council, NIH
Office for National Statistics). FINRISK: This study was supported by the Heart, Lung and Blood Institute and Economic and Social Research Council);
Academy of Finland Center of Excellence in Complex Disease Genetics (grant financial relationships with any organisations that might have an interest in
numbers 213506, 129680), the Academy of Finland (grant numbers 139635, the submitted work in the previous 3 years (LJH: Australian National Health
129494, 136895, 263836 and 141054), the Sigrid Juselius Foundation, and and Medical Research council, PMV: Swiss National Science Foundation, MP:
ENGAGE – European Network for Genetic and Genomic Epidemiology, Eli Lilly and Lundbeck advisory board membership, Swiss National Science
FP7-HEALTH-F4-2007, grant agreement number 201413. Generation Scotland: Foundation, TK: Pfizer consultancy on tobacco dependence, Juho Vainio
Generation Scotland has received core funding from the Chief Scientist Office Foundation, The Finnish Medical Society, TSA: Gene Diet Interactions in
of the Scottish Government Health Directorates CZD/16/6 and the Scottish Obesity project, HT: KOC University, Pfizer), YBS: grants from NIHR, MRC,
Funding Council HR03006. Genotyping of the GS:SFHS samples was carried PDUK, BHF. AET, MRM and MEF are members of the UK Centre for Tobacco
out by the Genetics Core Laboratory at the Wellcome Trust Clinical Research and Alcohol Studies, a UKCRC Public Health Research: Centre of Excellence.
Facility, Edinburgh, Scotland and was funded by the UK’s Medical Research Funding from British Heart Foundation, Cancer Research UK, Economic and
Council. GOYA females: The GOYA study was conducted as part of the Social Research Council, Medical Research Council, and the National Institute
activities of the Danish Obesity Research Centre (DanORC, www.danorc.dk) for Health Research, under the auspices of the UK Clinical Research
and The MRC centre for Causal Analyses in Translational Epidemiology (MRC Collaboration is gratefully acknowledged. LTL is supported by BHF
CAiTE). The genotyping for GOYA was funded by the Wellcome Trust (WT programme grant RG/08/013/25942. LP is funded by an MRC Population
084762). GOYA is a nested study within The Danish National Birth Cohort Health Scientist Fellowship (MR/J012165/1). TSA was supported by the Gene
which was established with major funding from the Danish National Research Diet Interactions in Obesity (GENDINOB, www.gendinob.dk) postdoctoral
Foundation. Additional support for this cohort has been obtained from the fellowship grant. LLNH was supported by the Health Insurance Foundation
Pharmacy Foundation, the Egmont Foundation, The March of Dimes Birth (grant No. 2010 B 131). The work of SSM was funded by the Netherlands
Defects Foundation, the Augustinus Foundation, and the Health Foundation. Genomics Initiative (NGI)/Netherlands Organization for Scientific Research
HBCS: Helsinki Birth Cohort Study has been supported by grants from the (NWO) (grant-numbers 050-060-810 NCHA); the work of HT was supported
Academy of Finland, the Finnish Diabetes Research Society, Folkhälsan by NWO-VIDI (grant 017-106-370). MeK is partially supported by the
Research Foundation, Novo Nordisk Foundation, Finska Läkaresällskapet, Economic and Social Research Council International Centre for Life Course
Signe and Ane Gyllenberg Foundation, University of Helsinki, Ministry of Studies in Society and Health (RES-596-28-0001).
Education, Ahokas Foundation, Emil Aaltonen Foundation. NFBC: NFBC1966
and NFBC1986 received financial support from the Academy of Finland Patient consent Obtained.
( project grants 104781, 120315, 129269, 1114194, 24300796, 141042 Ethics approval 1958BC: Ethics approval for the study was obtained from the
Center of Excellence in Complex Disease Genetics and SALVE), University South-East Multi-Centre Research Ethics Committee (Ref: 01/1/44) and the
Hospital Oulu, Biocenter, University of Oulu, Finland (75617), NHLBI grant Joint UCL/UCLH Committees on the Ethics of Human Research (Committee A)
5R01HL087679-02 through the STAMPEED program (1RL1MH083268-01), Ref: 08/H0714/40. ALSPAC: Ethics approval for the study was obtained from
NIH/NIMH (5R01MH63706:02), the European Commission (EURO-BLCS, the ALSPAC Ethics and Law Committee and the Local Research Ethics
Framework 5 award QLG1-CT-2000-01643), ENGAGE project and grant Committee. BRHS: The BRHS has local (from each of the districts in which
agreement HEALTH-F4-2007-201413, EU FP7 EurHEALTHAgeing -277849, the the study was based) and multicentre ethical committee approvals. Ethical
Medical Research Council, UK (G0500539, G0600705, G1002319, approval was granted for the BWHHS from the London Multi-Centre Research
PrevMetSyn/SALVE) and the MRC, Centenary Early Career Award. The DNA Ethics Committee and 23 Local Research Ethics Committees. BWHHS: Ethical
extractions, sample quality controls, biobank up-keeping and aliquotting was approval was granted for the BWHHS from the London Multi-Centre Research
performed in the National Public Health Institute, Biomedicum Helsinki, Ethics Committee and 23 Local Research Ethics Committees. CaPS: Ethics
Finland and supported financially by the Academy of Finland and Biocentrum approval was obtained from the South Glamorgan Area Health Authority, the
Helsinki. NSHD: This work was funded by the Medical Research Council Gwent REC, and the South Wales Research Ethics Committee D. CHDS: All
[MC_UU_12019/1]. NTR: This study was supported by the European phases of the study have received ethical approval from the regional Health
Research Council (ERC Starting Grant 284167 PI Vink), Netherlands and Disability Ethics Committee. Colaus/PsyCoLaus: Colaus and PsyCoLaus
Organization for Scientific Research (NWO: MagW/ZonMW grants were approved by the Institutional Ethic’s Committee of the University of
904-61-090, 985-10-002, 904-61-193, 480-04-004, 400-05-717, Lausanne. ELSA: ELSA has been approved by the National Research Ethics
Addiction-31160008 Middelgroot-911-09-032, Spinozapremie 56-464-14192), Service. FINRISK: The 2002 and 2007 FINRISK surveys have been approved
BBRMI-NL (Biobanking and Biomolecular Resources Research Infrastructure), by the Coordinating Ethics Committee of the Helsinki University Hospital
VU University’s Institutes for Health and Care Research and Neuroscience District. Generation Scotland: Ethical approval for the study was given by the
Campus Amsterdam. Patch 2: The Patch 2 study was funded by the Imperial NHS Tayside committee on research ethics (reference 05/s1401/89). GOYA
Cancer Research Fund (ICAF) and Cancer Research UK (CRUK) Programme females: The study was approved by the regional scientific ethics committee
Grants to the General Practice Research Group at the University of Oxford. and by the Danish Data Protection Board. HBCS: The research plan of the
Rotterdam: The Rotterdam Study was supported by the Erasmus Medical HBCS was approved by the Institutional Review Board of the National Public
Center and the Erasmus University Rotterdam, the Netherlands Organisation Health Institute. Health2006/Health2008/Inter99: The studies have been
of Scientific Research (NWO), the Netherlands Organisation for Health approved by the Ethical Committee of Copenhagen. HUNT: Use of data in the
Research and Development (ZonMw), the Ministry of Education, Culture, and present study was approved by the Regional Committee for Medical Research
Science, the Ministry of Health, Welfare, and Sports, the European Ethics (Reference nr. 2013/1127/REK midt). NFBC: The University of Oulu
Commission (DG-XII). SYS-P: The SYS-P is funded by the Canadian Institutes Ethics Committee and the Ethical Committee of Northern Ostrobothnia
of Health Research, the Canadian Foundation for Innovation, and the Heart Hospital District have approved the study. NHANES: Data collection for
and Stroke Foundation of Ontario. Whitehall: The Whitehall II study has been NHANES was approved by the NCHS Research Ethics Review Board. Analysis
supported by grants from the Medical Research Council; British Heart of de-identified data from the survey is exempt from the federal regulations
Foundation; Health and Safety Executive; Department of Health; National Heart for the protection of human research participants. Analysis of restricted data
Lung and Blood Institute (NHLBI: HL36310) and National Institute on Aging through the NCHS Research Data Center is also approved by the NCHS ERB.
(AG13196), US, NIH; Agency for Health Care Policy Research (HS06516); and NSHD: Ethical approval was given by the Central Manchester Research Ethics
the John D and Catherine T MacArthur Foundation Research Networks on Committee. NTR: The NTR study was approved by the Central Ethics
Successful Midlife Development and Socio-economic Status and Health. Committee on Research Involving Human Subjects of the VU University
Competing interests MRM: Medical Research Council, LTL: British Heart Medical Center, Amsterdam (IRB number IRB-2991 under Federalwide
Foundation, LJH: New Zealand Health Research Council, MK: Jim and Mary Assurance 3703; IRB/institute code 03-180). Patch 2: Ethical approval was
Carney Charitable Trust, New Zealand Health Research Council, CCM: obtained from the Anglia and Oxford Multicentre Research Ethics Committee
European Research Council, DIB, JMV: NWO, European Research Council, and from the Local Research Ethics Committees covering the areas of

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 11

 Open Access

residence of the patients. Rotterdam: The Medical Ethics Committee of saguenay-youth-study.org) (scientific community/collaborative interests).
Erasmus Medical Center Rotterdam approved the study. SYS-P: The Research Whitehall: Data from the Whitehall II study are made publicly available as
Ethics Committee of the Chicoutimi Hospital in Quebec, Canada approved the described in the Whitehall II data sharing policy (http://www.ucl.ac.uk/
study protocol. Whitehall: Ethical approval for the Whitehall II study was whitehallII/datasharing).
obtained from the University College London Medical School committee on
Open Access This is an Open Access article distributed in accordance with
the ethics of human research.
the terms of the Creative Commons Attribution (CC BY 4.0) license, which
Provenance and peer review Not commissioned; externally peer reviewed. permits others to distribute, remix, adapt and build upon this work, for
commercial use, provided the original work is properly cited. See: http://
Data sharing statement 1958BC: This study makes use of data generated by creativecommons.org/licenses/by/4.0/
the Wellcome Trust Case-Control Consortium. A full list of investigators who
contributed to generation of the data is available from the Wellcome Trust
Case-Control Consortium website. The 1958 birth cohort data can be REFERENCES
accessed via the UK Data Service (http://ukdataservice.ac.uk/). ALSPAC: Data 1. Weyerer S, Eifflaender-Gorfer S, Wiese B, et al. Incidence and
used for this submission will be made available on request to the ALSPAC predictors of depression in non-demented primary care attenders
executive committee (alspac-exec@bristol.ac.uk). The ALSPAC data aged 75 years and older: results from a 3-year follow-up study. Age
management plan (available here:http://www.bristol.ac.uk/alspac/researchers/ Ageing 2013;42:173–80.
data-access/) describes in detail the policy regarding data sharing, which is 2. Chaiton MO, Cohen JE, O’Loughlin J, et al. A systematic review of
longitudinal studies on the association between depression and
through a system of managed open access. BRHS: We welcome proposals for smoking in adolescents. BMC Public Health 2009;9:356.
collaborative projects and data sharing (http://www.ucl.ac.uk/pcph/research- 3. Audrain-McGovern J, Rodriguez D, Rodgers K, et al. Declining
groups-themes/brhs-pub). For general data sharing enquiries, please contact alternative reinforcers link depression to young adult smoking.
Lucy Lennon (l.lennon@ucl.ac.uk). BWHHS: All BWHHS data collected is held Addiction 2011;106:178–87.
by the research team based at London School of Hygiene and Tropical Medicine, 4. Franko DL, Striegel-Moore RH, Bean J, et al. Psychosocial and
health consequences of adolescent depression in Black and White
for ongoing analysis. If you would like to collaborate with the BWHHS team,
young adult women. Health Psychol 2005;24:586–93.
contact the study coordinator, AA (antoinette.amuzu@lshtm.ac.uk). Data and 5. Dierker L, Donny E. The role of psychiatric disorders in the
biological samples provided to the collaborators can only be used for the relationship between cigarette smoking and DSM-IV nicotine
purposes originally stated and must not be used in any other way without dependence among young adults. Nicotine Tob Res 2008;10:439–46.
re-application to the BWHHS team. No data should be passed on to any third 6. Byers AL, Vittinghoff E, Lui LY, et al. Twenty-year depressive
party unless they were specified in the original application. CaPS: Data used trajectories among older women. Arch Gen Psychiatry
2012;69:1073–9.
for the Caerphilly Prospective study (CaPS) was made available by the CaPS 7. Leung J, Gartner C, Hall W, et al. A longitudinal study of the
access committee (Chair: Professor Kay Tee Khaw). More infomation about its bi-directional relationship between tobacco smoking and
managed access procedure is available on the study website (http://www.bris. psychological distress in a community sample of young Australian
ac.uk/social-community-medicine/people/project/1392). CHDS: Data women. Psychol Med 2012;42:1273–82.
contributed for this submission are available on request from the CHDS 8. Tjora T, Hetland J, Aaro LE, et al. The association between smoking
( john.horwood@otago.ac.nz). Colaus/PsyCoLaus: Data from the CoLaus/ and depression from adolescence to adulthood. Addiction
2014;109:1022–30.
PsyCoLaus study can be requested according to the procedure described on 9. Moylan S, Jacka FN, Pasco JA, et al. Cigarette smoking, nicotine
the CoLaus website (http://www.colaus.ch/en/cls_home/cls_pro_home/ dependence and anxiety disorders: a systematic review of
cls-research-3.htm). ELSA: ELSA data are made available through the ESDS population-based, epidemiological studies. BMC Med 2012;10:123.
website (http://www.elsa-project.ac.uk/availableData). FINRISK: Data used for 10. Cook BL, Wayne GF, Kafali EN, et al. Trends in smoking among
this submission will be made available on request to the FINRISK adults with mental illness and association between mental health
treatment and smoking cessation. JAMA 2014;311:172–82.
Management Group, according to the given ethical guidelines and Finnish
11. Warren CW, Jones NR, Eriksen MP, et al. Patterns of global tobacco
legislation. Generation Scotland: Data is available on request (access@ use in young people and implications for future chronic disease
generationscotland.org). GOYA females: An anonymized copy of the data used burden in adults. Lancet 2006;367:749–53.
for this submission will be made available on request to the GOYA analysts 12. Ferrari AJ, Charlson FJ, Norman RE, et al. Burden of depressive
after permission have been given by the DNBC executive committee disorders by country, sex, age, and year: findings from the global
(www.dnbc.dk). HBCS: Data used for this submission will be made available burden of disease study 2010. PLoS Med 2013;10:e1001547.
13. Gage SH, Davey Smith G, Zammit S, et al. Using Mendelian
on request to the HBCS executive committee ( johan.eriksson@helsinki.fi). randomisation to infer causality in depression and anxiety research.
Health2006/Health2008/Inter99: Data used for this submission can be made Depress Anxiety 2013;30:1185–93.
available on request to the Research Centre for Prevention and Health 14. Munafo MR, Araya R. Cigarette smoking and depression: a question
(http://www.regionh.dk/fcfs/Menu/). Please contact LLNH (lise.lotte.nystrup. of causation. Br J Psychiatry 2010;196:425–6.
husemoen@regionh.dk) or AL (allan.linneberg@regionh.dk). HUNT: Data used 15. Dierker LC, Avenevoli S, Merikangas KR, et al. Association between
from the HUNT Study for this submission will be made available on request psychiatric disorders and the progression of tobacco use behaviors.
J Am Acad Child Adolesc Psychiatry 2001;40:1159–67.
to the HUNT Data Access Committee (hunt@medisin.ntnu.no). The HUNT data 16. Breslau N, Peterson EL, Schultz LR, et al. Major depression and
access information (http://www.ntnu.edu/hunt/data) describes in detail the stages of smoking. A longitudinal investigation. Arch Gen Psychiatry
policy regarding data availability. NFBC: Data used for this submission can be 1998;55:161–6.
made available on request to Tuula Ylitalo (tuula.ylitalo@oulu.fi), Minna 17. Patton GC, Carlin JB, Coffey C, et al. Depression, anxiety, and
Mannikko (minna.annikko@oulu.fi) or M-RJ (m.jarvelin@imperial.ac.uk). smoking initiation: a prospective study over 3 years. Am J Public
Health 1998;88:1518–22.
NHANES: NHANES data can be accessed here: (http://www.cdc.gov/nchs/
18. Burgess ES, Brown RA, Kahler CW, et al. Patterns of change in
nhanes.htm). The genotype used in this analysis is a restricted variable. depressive symptoms during smoking cessation: Who’s at risk for
Applications for access to these data must be made through the Research relapse? J Consult Clin Psychol 2002;70:356–61.
Data Center: (http://www.cdc.gov/rdc/). NSHD: The NSHD data are made 19. Munafo MR, Hitsman B, Rende R, et al. Effects of progression to
available to researchers who submit data requests (tomrclha.swiftinfo@ucl.ac. cigarette smoking on depressed mood in adolescents: evidence
uk). More information is available in the full policy documents (http://www. from the National Longitudinal Study of Adolescent Health. Addiction
2008;103:162–71.
nshd.mrc.ac.uk/data.aspx). Managed access is in place for this study to 20. Boden JM, Fergusson DM, Horwood LJ. Cigarette smoking and
ensure that use of the data are within the bounds of consent given previously depression: tests of causal linkages using a longitudinal birth cohort.
by participants, and to safeguard any potential threat to anonymity since the Br J Psychiatry 2010;196:440–6.
participants are all born in the same week. NTR: Data used for this 21. Breslau N, Novak SP, Kessler RC. Daily smoking and the subsequent
submission will be made available on request to the NTR committee onset of psychiatric disorders. Psychol Med 2004;34:323–33.
(ntr@psy.vu.nl). SYS-P: Data used for this submission will be made available 22. Taylor G, McNeill A, Girling A, et al. Change in mental health after
smoking cessation: systematic review and meta-analysis. BMJ
on request to the SYS principal investigators (http://www. 2014;348:g1151.

12 Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141

 Open Access

23. Markou A, Kosten TR, Koob GF. Neurobiological similarities in 46. Moon SS, Mo BC, Basham R. Adolescent depression and future
depression and drug dependence: a self-medication hypothesis. smoking behavior: a prospective study. Child Adolesc Soc Work J
Neuropsychopharmacology 1998;18:135–74. 2010;27:405–22.
24. Badrick E, Kirschbaum C, Kumari M. The relationship between 47. Escobedo LG, Reddy M, Giovino GA. The relationship between
smoking status and cortisol secretion. J Clin Endocrinol Metab depressive symptoms and cigarette smoking in US adolescents.
2007;92:819–24. Addiction 1998;93:433–40.
25. Rose JE, Behm FM, Ramsey C, et al. Platelet monoamine oxidase, 48. McKenzie M, Olsson CA, Jorm AF, et al. Association of adolescent
smoking cessation, and tobacco withdrawal symptoms. Nicotine Tob symptoms of depression and anxiety with daily smoking and nicotine
Res 2001;3:383–90. dependence in young adulthood: findings from a 10-year longitudinal
26. Saal D, Dong Y, Bonci A, et al. Drugs of abuse and stress trigger a study. Addiction 2010;105:1652–9.
common synaptic adaptation in dopamine neurons. Neuron 49. Ripke S, Wray NR, Lewis CM, et al. Major Depressive Disorder
2003;37:577–82. Working Group of the Psychiatric GC. A mega-analysis of
27. Davey Smith G, Ebrahim S. ‘Mendelian randomization’: can genetic genome-wide association studies for major depressive disorder.
epidemiology contribute to understanding environmental Mol Psychiatry 2013;18:497–511.
determinants of disease? Int J Epidemiol 2003;32:1–22. 50. Taylor AE, Davies NM, Ware JJ, et al. Mendelian randomization in
28. Davey Smith G. Use of genetic markers and gene-diet interactions health research: using appropriate genetic variants and avoiding
for interrogating population-level causal influences of diet on health. biased estimates. Econ Hum Biol 2013;13:99–106.
Genes Nutr 2011;6:27–43. 51. VanderWeele TJ, Tchetgen Tchetgen EJ, Cornelis M, et al.
29. Furberg H, Kim Y, Dackor J, et al. Genome-wide meta-analyses Methodological challenges in Mendelian randomization.
identify multiple loci associated with smoking behavior. Nat Genet Epidemiology 2014;25:427–35.
2010;42:441–7. 52. Hill AB. The environment and disease: association or causation?
30. Ware JJ, van den Bree M, Munafo MR. From men to mice: Proc R Soc Med 1965;58:295–300.
CHRNA5/CHRNA3, smoking behavior and disease. Nicotine Tob 53. Loukola A, Hallfors J, Korhonen T, et al. Genetics and smoking.
Res 2012;14:1291–9. Curr Addict Rep 2014;1:75–82.
31. Munafo MR, Timofeeva MN, Morris RW, et al. Association between 54. Li MD, Cheng R, Ma JZ, et al. A meta-analysis of estimated genetic
genetic variants on chromosome 15q25 locus and objective and environmental effects on smoking behavior in male and female
measures of tobacco exposure. J Natl Cancer Inst 2012;104:740–8. adult twins. Addiction 2003;98:23–31.
32. Ducci F, Kaakinen M, Pouta A, et al. TTC12-ANKK1-DRD2 and 55. Saccone NL, Culverhouse RC, Schwantes-An TH, et al. Multiple
CHRNA5-CHRNA3-CHRNB4 influence different pathways leading to independent loci at chromosome 15q25.1 affect smoking quantity:
smoking behavior from adolescence to mid-adulthood. Biol a meta-analysis and comparison with lung cancer and COPD.
Psychiatry 2011;69:650–60. PLoS Genet 2010;6:pii: e1001053.
33. Fowler CD, Lu Q, Johnson PM, et al. Habenular alpha5 nicotinic 56. Thorgeirsson TE, Gudbjartsson DF, Surakka I, et al. Sequence
receptor subunit signalling controls nicotine intake. Nature variants at CHRNB3-CHRNA6 and CYP2A6 affect smoking
2011;471:597–601. behavior. Nat Genet 2010;42:448–53.
34. Ware JJ, van den Bree MB, Munafo MR. Association of the 57. Picciotto MR, Brunzell DH, Caldarone BJ. Effect of nicotine and
CHRNA5-A3-B4 gene cluster with heaviness of smoking: nicotinic receptors on anxiety and depression. Neuroreport
a meta-analysis. Nicotine Tob Res 2011;13:1167–75. 2002;13:1097–106.
35. Keskitalo K, Broms U, Heliovaara M, et al. Association of serum 58. Otowa T, Kawamura Y, Nishida N, et al. Meta-analysis of
cotinine level with a cluster of three nicotinic acetylcholine receptor genome-wide association studies for panic disorder in the Japanese
genes (CHRNA3/CHRNA5/CHRNB4) on chromosome 15. Hum Mol population. Transl Psychiatry 2012;2:e186.
Genet 2009;18:4007–12. 59. Trzaskowski M, Eley TC, Davis OS, et al. First genome-wide
36. Freathy RM, Ring SM, Shields B, et al. A common genetic variant in association study on anxiety-related behaviours in childhood. PLoS
the 15q24 nicotinic acetylcholine receptor gene cluster ONE 2013;8:e58676.
(CHRNA5-CHRNA3-CHRNB4) is associated with a reduced ability 60. Glymour MM, Tchetgen EJ, Robins JM. Credible Mendelian
of women to quit smoking in pregnancy. Hum Mol Genet randomization studies: approaches for evaluating the instrumental
2009;18:2922–7. variable assumptions. Am J Epidemiol 2012;175:332–9.
37. Asvold BO, Bjørngaard JH, Carslake D, et al. Causal associations of 61. Cole SR, Platt RW, Schisterman EF, et al. Illustrating bias due to
tobacco smoking with cardiovascular risk factors: a Mendelian conditioning on a collider. Int J Epidemiol 2010;39:417–20.
randomization analysis of the HUNT Study in Norway. Int J 62. Munafo MR, Johnstone EC, Walther D, et al. CHRNA3 rs1051730
Epidemiol 2014 May 26. pii: dyu113. [Epub ahead of print] PubMed genotype and short-term smoking cessation. Nicotine Tob Res
PMID: 24867305. 2011;13:982–8.
38. Freathy RM, Kazeem GR, Morris RW, et al. Genetic variation at 63. Bergen AW, Javitz HS, Krasnow R, et al. Nicotinic acetylcholine
CHRNA5-CHRNA3-CHRNB4 interacts with smoking status to receptor variation and response to smoking cessation therapies.
influence body mass index. Int J Epidemiol 2011;40:1617–28. Pharmacogenet Genomics 2013;23:94–103.
39. Tyrrell J, Huikari V, Christie JT, et al. Genetic variation in the 15q25 64. Copeland KT, Checkoway H, McMichael AJ, et al. Bias due to
nicotinic acetylcholine receptor gene cluster (CHRNA5-CHRNA3- misclassification in the estimation of relative risk. Am J Epidemiol
CHRNB4) interacts with maternal self-reported smoking status 1977;105:488–95.
during pregnancy to influence birth weight. Hum Mol Genet 65. Johansson R, Carlbring P, Heedman A, et al. Depression, anxiety
2012;21:5344–58. and their comorbidity in the Swedish general population: point
40. Bjorngaard JH, Gunnell D, Elvestad MB, et al. The causal role of prevalence and the effect on health-related quality of life. Peer J
smoking in anxiety and depression: a Mendelian randomization 2013;1:e98.
analysis of the HUNT study. Psychol Med 2013;43:711–19. 66. Judd LL, Kessler RC, Paulus MP, et al. Comorbidity as a
41. Lewis SJ, Araya R, Smith GD, et al. Smoking is associated with, but fundamental feature of generalized anxiety disorders: results from
does not cause, depressed mood in pregnancy—a Mendelian the National Comorbidity Study (NCS). Acta Psychiatr Scand Suppl
randomization study. PLoS ONE 2011;6:e21689. 1998;393:6–11.
42. Needham BL. Gender differences in trajectories of depressive 67. Sandanger I, Moum T, Ingebrigtsen G, et al. Concordance between
symptomatology and substance use during the transition from symptom screening and diagnostic procedure: the Hopkins
adolescence to young adulthood. Soc Sci Med 2007;65:1166–79. Symptom Checklist-25 and the Composite International
43. Fleming CB, Mason WA, Mazza JJ, et al. Latent growth modeling of Diagnostic Interview I. Soc Psychiatry Psychiatr Epidemiol
the relationship between depressive symptoms and substance use 1998;33:345–54.
during adolescence. Psychol Addict Behav 2008;22:186–97. 68. Head J, Stansfeld SA, Ebmeier KP, et al. Use of self-administered
44. Breslau N, Kilbey MM, Andreski P. Nicotine dependence and major instruments to assess psychiatric disorders in older people:
depression. New evidence from a prospective investigation. Arch validity of the General Health Questionnaire, the Center for
Gen Psychiatry 1993;50:31–5. Epidemiologic Studies Depression Scale and the self-completion
45. Kendler KS, Neale MC, MacLean CJ, et al. Smoking and major version of the revised Clinical Interview Schedule. Psychol Med
depression. A causal analysis. Arch Gen Psychiatry 1993;50:36–43. 2013;43:2649–56.

Taylor AE, et al. BMJ Open 2014;4:e006141. doi:10.1136/bmjopen-2014-006141 13