Content uploaded by Carsten Hjorthøj
Author content
All content in this area was uploaded by Carsten Hjorthøj on Jan 04, 2016
Content may be subject to copyright.
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
1
Articles
Association between alcohol and substance use disorders
and all-cause and cause-specifi c mortality in schizophrenia,
bipolar disorder, and unipolar depression: a nationwide,
prospective, register-based study
Carsten Hjorthøj, Marie Louise Drivsholm Østergaard, Michael Eriksen Benros, Nanna Gilliam Toftdahl, Annette Erlangsen, Jon Trærup Andersen,
Merete Nordentoft
Summary
Background People with severe mental illness have both increased mortality and are more likely to have a substance
use disorder. We assessed the association between mortality and lifetime substance use disorder in patients with
schizophrenia, bipolar disorder, or unipolar depression.
Methods In this prospective, register-based cohort study, we obtained data for all people with schizophrenia, bipolar
disorder, or unipolar depression born in Denmark in 1955 or later from linked nationwide registers. We obtained
information about treatment for substance use disorders (categorised into treatment for alcohol, cannabis, or hard
drug misuse), date of death, primary cause of death, and education level. We calculated hazard ratios (HRs) for
all-cause mortality and subhazard ratios (SHRs) for cause-specifi c mortality associated with substance use disorder of
alcohol, cannabis, or hard drugs. We calculated standardised mortality ratios (SMRs) to compare the mortality in the
study populations to that of the background population.
Findings Our population included 41 470 people with schizophrenia, 11 739 people with bipolar disorder, and 88 270 people
with depression. In schizophrenia, the SMR in those with lifetime substance use disorder was 8·46 (95% CI 8·14–8·79),
compared with 3·63 (3·42–3·83) in those without. The respective SMRs in bipolar disorder were 6·47 (5·87–7·06) and
2·93 (2·56–3·29), and in depression were 6·08 (5·82–6·34) and 1·93 (1·82–2·05). In schizophrenia, all substance use
disorders were signifi cantly associated with increased risk of all-cause mortality, both individually (alcohol, HR 1·52
[95% CI 1·40–1·65], p<0·0001; cannabis, 1·24 [1·04–1·48], p=0·0174; hard drugs, 1·78 [1·56–2·04], p<0·0001) and when
combined. In bipolar disorder or depression, only substance use disorders of alcohol (bipolar disorder, HR 1·52 [95% CI
1·27–1·81], p<0·0001; depression, 2·01 [1·86–2·18], p<0·0001) or hard drugs (bipolar disorder, 1·89 [1·34–2·66],
p=0·0003; depression, 2·27 [1·98–2·60], p<0·0001) increased risk of all-cause mortality individually.
Interpretation Mortality in people with mental illness is far higher in individuals with substance use disorders than in
those without, particularly in people who misuse alcohol and hard drugs. Mortality-reducing interventions should
focus on patients with a dual diagnosis and seek to prevent or treat substance use disorders.
Funding The Lundbeck Foundation.
Introduction
Severe mental illness has been linked to excess mortality
from suicide, accidents, and somatic illnesses.1–3 Similarly,
misuse or dependence on alcohol or illicit substances is
associated with increased mortality.4–6 Substance use
disorders are prevalent in people with severe mental
illness, and might be present in as many as half of people
with schizophrenia.7–10 Little is known about the mortality
of people diagnosed with both mental illness and substance
use disorders. Results of a study of 762 patients with
psychotic disorders showed that cannabis was associated
with a reduction in mortality by more than half, but alcohol
had no eff ect on mortality.11 Furthermore, studies have
linked cannabis use to lung cancer independently of
tobacco smoking.12,13
A better understanding of the potential contribution of
substance use disorders to the raised mortality in people
with mental illness will provide clues to reduce this excess
mortality in these patients. A reduced lifespan by 15–20 years
in people with mental illness has been previously described,
but without adjustment for sub stance use.3 Ultimately, this
information can guide prevention, detection, and treatment
programmes for substance use disorders.
Our aim of this study was to assess mortality in a large
cohort of patients with severe mental illness in relation
to lifetime diagnosis of substance use disorder. We aimed
to establish specifi c substance-related risk factors in
people with schizophrenia, bipolar disorder, and unipolar
depression.
Methods
Study design and population
In this prospective, register-based study, we linked
nationwide Danish registers using the Danish Civil
Lancet Psychiatry 2015
Published Online
August 13, 2015
http://dx.doi.org/10.1016/
S2215-0366(15)00207-2
See Online/Comment
http://dx.doi.org/10.1016/
S2215-0366(15)00372-7
Mental Health Center,
Copenhagen University
Hospital, Copenhagen,
Denmark (C Hjorthøj PhD,
M L D Østergaard MSc,
M E Benros PhD,
N G Toftdahl BScMed,
A Erlangsen PhD,
Prof M Nordentoft DrMedSc);
The Lundbeck Foundation
Initiative for Integrative
Psychiatric Research, iPSYCH,
Aarhus and Copenhagen,
Denmark (C Hjorthøj,
M L D Østergaard, M E Benros,
N G Toftdahl,
Prof M Nordentoft);
Department of Mental Health,
Johns Hopkins School of Public
Health, Baltimore, MD, USA
(A Erlangsen); Laboratory of
Clinical Pharmacology Q7642,
Rigshospitalet, Copenhagen
University Hospital,
Copenhagen, Denmark
(J T Andersen PhD); and
Department of Clinical
Pharmacology, Bispebjerg
Hospital, Copenhagen,
Denmark (J T Andersen)
Correspondence to:
Dr Carsten Hjorthøj, Copenhagen
University Hospital, Mental
Health Center, Copenhagen,
Kildegårdsvej 28, opgang 15.4,
DK-2900 Hellerup, Denmark
carsten.hjorthoej@regionh.dk
Articles
2
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
Registration System, which has provided all permanent
residents in Denmark since 1968 with a unique
identifi cation number that enables linkage of individuals
across registers.14 An overview of the registers used is
presented in table 1.
We used the Psychiatric Central Research Register15 to
defi ne three study populations: patients diagnosed within
the schizophrenia spectrum (henceforth referred to as
schizophrenia), bipolar disorder, or unipolar depression
(henceforth referred to as depression). This register
contains information about all psychiatric admissions
since 1969, and outpatient visits since 1995. The diagnostic
codes used to defi ne the disorder categories are shown in
the appendix. We used data for all individuals born in
Denmark on or after Jan 1, 1955, only. We chose this date
so that we could be certain that the fi rst registered contact
with psychiatric illness was the incident illness; patients
born in 1955 would be aged 14 years when the register
was established, and presentation of these disorders
before age 14 years would have been unlikely.
Register-based analyses in Denmark do not require
participant consent. The protocol was approved by the
Danish Data Protection Agency.
Procedures
We categorised information about treatment for
substance use disorders into treatment for alcohol,
cannabis, and hard-drug misuse (ie, opioids, cocaine,
stimulants, sedatives, hallucinogens, and volatile
solvents) using diagnostic codes and anatomical
therapeutic chemical (ATC) classifi cation system codes
(appendix). We combined data from the Psychiatric
Central Research Register, National Patient Register,16
National Prescription Registry,17 and Registry of Drug
Abusers Undergoing Treatment.18 We also obtained
information about use of prescription medicines used to
treat alcohol and hard-drug misuse—eg, disulfi ram for
alcohol use and methadone for opioid dependence.
We obtained information about substance use and other
treatment until April, 2013. We obtained information
about date of death from the Central Persons Registry
until July 17, 2013, and about the primary cause of death
established by a medical doctor from the Cause of Death
Registry until Dec 31, 2011.19 Cause-specifi c deaths were
censored on Dec 31, 2011, because the register containing
the information was not updated. We used information
from Statistics Denmark to obtain information about the
highest level of education (either less than high school,
high school, vocational training, or college).20
Statistical analysis
We used Cox regression (hazard ratios [HRs]) to analyse
all-cause mortality. For cause-specifi c mortality, we
estimated subhazard ratios (SHRs) according to the
method of Fine and Gray.21 The individuals were followed
Research in context
Evidence before this study
We searched MEDLINE for combinations of the search terms
“schizophrenia”, “bipolar”, “unipolar”, “depression”, “severe
mental illness”, and “death”, “mortality”, “suicide”, “accidents”,
and “alcohol”, “cannabis”, “illicit substances”, “cocaine”,
“heroin”, “opioids”, “amphetamine”. We applied no language or
publication date restrictions, and we did our last search in
May, 2014. We focused on epidemiological studies examining
the mortality (all-cause or cause-specifi c) related to substance
use in schizophrenia, bipolar disorder, or unipolar depression.
Evidence from the studies that we found showed that severe
mental illness is strongly associated with increased mortality,
and that people with severe mental illness are also much more
likely to misuse alcohol, cannabis, and other illicit substances.
None of the studies that we found had sought to investigate
how much of the increased mortality in severe mental illness
can be explained by substance use disorders.
Added value of this study
As expected, all substances increased mortality in all
three populations. The estimated proportion of mortality
attributable to substances, and the size of the individual
associations, have not been reported in previous studies.
Implications of the available evidence
Although our results cannot prove causality, the prevention and
treatment of misuse of alcohol, cannabis, and other illicit
substances could help to reduce the excess mortality
consistently reported in people with severe mental illness.
People with dual diagnosis remain an important group for
whom to focus mortality prevention.
See Online for appendix
Data Period covered
Danish Civil Registration System Unique personal identifi cation number; sex;
date of birth; date of death; date of emigration
Since 1968
Cause of Death Registry Dates and causes of death established by a
medical doctor
1970–2011 (inclusive)
Psychiatric Central Research
Register
Date of admission and discharge from
psychiatric hospitals (inpatient and outpatient)
along with diagnostic information; used both
to defi ne the study populations and determine
presence of substance use disorders
Since 1969
(outpatient data from
1995)
National Patient Register Date of admission and discharge from somatic
hospitals (inpatient and outpatient) along with
diagnostic information; used to determine
presence of substance use disorders
Since 1977
(outpatient data from
1995)
National Prescription Registry Redeemed prescriptions from pharmacies,
including dates and anatomical therapeutic
chemical (ATC) classifi cation system codes
Since 1995
Registry of Drug Abusers
Undergoing Treatment
Date of admission and primary substance used,
at specialised addiction facilities
Since 1996
Statistics Denmark Highest level of education Entire cohort
Table 1: Summary of registers used in the study
Articles
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
3
up from fi rst diagnosis until death, emigration, or end of
registers, whichever came fi rst. For cause-specifi c
deaths, individuals were censored when they died from
other causes. We adjusted all analyses for sex, birth year
(continuous variable), calendar year (continuous,
time-varying covariate), education, and age at onset of
the mental illness in question (continuous variable). We
do not present estimates and p values for covariates. For
all-cause mortality, suicide, and accidents, we
constructed a variable with separate categories for
substance use disorder of cannabis, alcohol, or hard
drugs, and combinations thereof. For the remaining
analyses, we created two models. In the fi rst model,
mortality risks of substance use disorder of cannabis,
alcohol, and hard drugs were estimated without mutual
adjustment. In the second model, each type of substance
use disorder was adjusted for the two others. Plots of
Schoenfeld residuals were used to assess the assumption
of proportional hazards. We did not correct p values for
multiple comparisons to avoid type II errors. Therefore,
p values in the range of 0·001 to 0·05 should be
interpreted with caution.
The fi rst date of substance use as recorded in the registers
(ie, the date that treatment was started) is very unlikely to
be the actual starting point of substance use, because the
patient is almost certainly going to have developed the
substance use disorder some time before starting
treatment. For this reason, we regarded people as exposed
to substance use disorders for the entire follow-up, even if
the fi rst diagnosis only occurred at a later timepoint. In
secondary analyses, we instead defi ned substance use
disorders as time-varying covariates. The eff ect and biases
of each of these approaches is discussed later.
We present the comparison of the study population to
the background population as age-and-sex-specifi c
standardised mortality ratios (SMRs)—ie, the observed
number of deaths divided by the expected number of
deaths, using aggregate data on the background population
available through Statistics Denmark for the period
1990–2012. We did all analyses using Stata/MP 13·1.
Schizophrenia (n=41 470) Bipolar disorder (n=11 739) Depression (n=88 270)
Age at fi rst diagnosis, years
Mean (SD) 27·6 (9·2) 34·0 (10·3) 31·3 (10·6)
Median (IQR) 25·7 (20·6–33·4) 33·8 (25·6–42·0) 30·7 (22·5–39·5)
Male individuals 24 127 (58·2%) 5092 (43·4%) 33 127 (37·5%)
Cannabis use disorder 7222 (17·4%) 1205 (10·3%) 5339 (6·0%)
Age at fi rst diagnosis of mental disorder, years 26·7 (8·3) 29·3 (9·3) 27·7 (8·9)
Alcohol use disorder 13 141 (31·7%) 3926 (33·4%) 21 014 (23·8%)
Age at fi rst diagnosis of mental disorder, years 29·7 (9·1) 33·4 (9·6) 32·0 (10·1)
Hard-drugs use disorder 9186 (22·2%) 2010 (17·1%) 10 659 (12·1%)
Age at fi rst diagnosis of mental disorder, years 28·9 (9·2) 33·3 (10·1) 32·5 (10·4)
Any substance use disorder 18 561 (44·8%) 4940 (42·1%) 27 428 (31·1%)
Substance use disorder predates mental disorder
by more than 1 year
7514 (40·5%) 2760 (55·9%) 12 625 (46·0%)
Substance use disorder and mental disorder fi rst
diagnosed within 1 year of one another
5764 (31·1%) 1204 (24·4%) 8590 (31·3%)
Mental disorder predates substance use disorder
by more than 1 year
5283 (28·5%) 976 (19·8%) 6213 (22·7%)
Person-years followed up 494 710 102 274 680 923
Deaths during follow-up 4616 (11·1%) 856 (7·3%) 3944 (4·5%)
Age at death, years
Mean (SD) 39·4 (9·6) 42·3 (9·2) 42·6 (9·3)
Median (IQR) 40·0 (32·0–47·4) 44·0 (36·0–49·7) 44·2 (36·3–49·8)
Highest level of education
Less than high school 23 036 (55·5%) 4290 (36·5%) 35 969 (40·7%)
High school 3811 (9·2%) 1121 (9·5%) 6744 (7·6%)
Vocational training 8214 (19·8%) 3419 (29·1%) 26 479 (30·0%)
College or university 4780 (11·5%) 2693 (22·9%) 17 184 (19·5%)
Not registered 1629 (3·9%) 216 (1·8%) 1894 (2·1%)
Lifetime diagnosis of schizophrenia 41 470 (100%) 3758 (32·0%) 11 217 (12·7%)
Lifetime diagnosis of bipolar disorder 3758 (9·1%) 11 739 (100%) 5741 (6·5%)
Lifetime diagnosis of depression 11 217 (27·0%) 5741 (48·9%) 88 270 (100%)
Data are n (%) unless otherwise specifi ed. Data are lifetime prevalence (while under observation).
Table 2: Characteristics of study population
Articles
4
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
Role of the funding source
The funder of the study had no role in study design, data
collection, data analysis, data interpretation, or writing of
the report. The corresponding author had full access to
all the data in the study and had fi nal responsibility for
the decision to submit for publication.
Results
Our population consisted of 41 470 people with
schizophrenia, 11 739 people with bipolar disorder, and
88 270 people with depression (table 2). 10 286 (24·8%)
people with schizophrenia, 3189 (27·2%) people with
bipolar disorder, and 19 540 (22·1%) people with
depression had one type of substance use disorder and
8275 (20·0%), 1751 (14·9%), and 7888 (8·9%), respectively,
had two or more types. Substance use disorders were
recorded in more than one register in 10 528 (56·7%) of
18 561 individuals with schizophrenia, 2948 (59·7%) of
4940 individuals with bipolar disorder, and 14 848 (54·1%)
of 27 428 individuals with depression. The proportion of
patients with a substance use disorder who were male
was higher than the proportion of patients without a
substance use disorder who were male: in schizophrenia,
12 723 (68·5%) of 18 561 of those with substance use
disorders were male vs 11 404 (49·8%) of 22 909 of
those without substance use disorder (p<0·0001). In
bipolar disorder, the corresponding proportions were
2637 (53·4%) of 4940 vs 2455 (36·1%) of 6799 (p<0·0001),
and in unipolar depression were 14 245 (51·9%) of
27 428 vs 18 882 (31·0%) of 60 842 (p<0·0001).
People with a lifetime substance use disorder of alcohol
or hard drugs had signifi cantly later onset of schizophrenia,
bipolar disorder, or depression (table 2) than did those
without substance use disorder, whose mean age at fi rst
diagnosis was 26·9 years (SD 9·4) for schizophrenia,
33·0 years (10·3) for bipolar disorder, and 30·4 years (10·6)
for depression (all p<0·0001, except p=0·0037 for hard
drugs in bipolar disorder). Individuals with cannabis use
disorder had an earlier onset of mental illness (all p<0·0001)
than did those without cannabis use disorder. In people
with schizophrenia and substance use disorder,
6997 (37·7%) had their fi rst substance disorder diagnosed
before incident schizophrenia, 1871 (10·1%) had the same
date registered for the two, and 9693 (52·2%) had their
fi rst substance use disorder diagnosed after incident
schizophrenia. In people with bipolar disorder and
substance use disorder, the corresponding fi gures were
1322 (26·8%), 323 (6·5%), and 3297 (66·7%), respectively,
whereas in depression, the corresponding fi gures were
9157 (33·4%), 2653 (9·7%), and 15 618 (56·9%).
For people with schizophrenia, the SMR of those with
a lifetime substance use disorder was 8·46 (95% CI
8·14–8·79), compared with 3·63 (3·42–3·83) in those
without (fi gure). For bipolar disorder, the corresponding
SMRs were 6·47 (5·87–7·06) and 2·93 (2·56–3·29),
respectively. For depression, the respective SMRs were
6·08 (5·82–6·34) and 1·93 (1·82–2·05; fi gure).
Substance use disorders of alcohol or hard drugs were
both linked to excess all-cause mortality in all three
psychiatric disorders (table 3). Only in schizophrenia was
cannabis use disorder by itself linked to excess mortality.
Alcohol misuse was associated with increased mortality
in all three disorders, with HRs ranging from 1·5 to 2,
whereas the increase was slightly higher for hard drugs
(HRs ranging from 1·8 to 2·3). Generally, combinations
of several types of substance use disorders increased
mortality additively. Three-way interaction terms were
never statistically signifi cant (data not shown). Two-way
interactions were signifi cant between alcohol and
cannabis (p=0·0038) and cannabis and hard drugs
(p=0·0223) in people with schizophrenia, and for alcohol
and hard drugs in people with depression (p<0·0001). All
signifi cant interactions suggested that the combined
eff ect was less than the sum of the individual substances.
Overall, suicide accounted for 1138 (25%) of 4616 deaths
in people with schizophrenia, 251 (29%) of 856 deaths in
people with bipolar disorder, and 998 (25%) of 3944 deaths
in people with depression. For schizophrenia, people
with exclusive alcohol use disorder (p=0·0141) or all three
types of substance use disorders (p=0·0021) were at a
lower risk of dying by suicide than were people with no
diagnosed substance use disorders (table 3). Schoenfeld
residuals suggested deviations from proportional hazards
for the analysis on alcohol use disorders, so this result
should be interpreted cautiously. For depression, people
with exclusive cannabis use disorder had a lower risk of
dying by suicide than did those with no substance use
disorders (p=0·0050), whereas combined substance use
disorder for both alcohol and hard drugs was linked to an
increased risk of death by suicide (p=0·0061; table 3). All
interactions were non-signifi cant (data not shown).
Deaths from accidents accounted for 881 (19%) deaths in
people with schizophrenia, 119 (14%) deaths in people with
bipolar disorder, and 562 (14%) deaths in people with
depression. Substance use disorder of alcohol or hard
drugs signifi cantly increased the risk of dying from
Men Women
0
1
2
4
6
8
10
Standardised mortality ratio
Schizophrenia
Bipolar disorder
Unipolar depression
Schizophrenia
Bipolar disorder
Unipolar depression
No substance use disorder
Substance use disorder
Figure: Standardised mortality ratio for the period 1990–2012 in people
with schizophrenia, bipolar disorder, or depression, with and without
substance use disorder
Error bars show 95% CI.
Articles
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
5
accidents, both combined and independently (table 3).
Cannabis use disorder only signifi cantly increased the risk
of death from accidents in schizophrenia. In people with
depression, adding cannabis use disorder to other types of
substance use disorders further increased the risk of death
from accidents, but this tendency was not noted in people
with bipolar disorder. The only statistically signifi cant inter-
actions were cannabis and hard drugs in schizophrenia
(p=0·0072) and alcohol and hard drugs in depression
(p=0·0001). Both interactions suggested that the combined
eff ect was less than the sum of the individual substances.
In our analysis of the risk of dying from one of four
major natural causes of death, we noted that 215 (5%)
deaths in people with schizophrenia and 264 (7%) deaths
in people with depression were caused by malignant
tumours outside the respiratory organs. We did not
analyse data for people with bipolar disorder because of
an insuffi cient number of deaths. In schizophrenia, no
substance use disorders were linked to increased
mortality from malignant tumours. In depression,
hard-drug use disorder was signifi cantly associated with
increased risk of dying from malignant tumours (table 4).
Cardiovascular causes accounted for 220 (5%) deaths in
people with schizophrenia and 152 (4%) deaths in those with
depression. In schizophrenia, alcohol misuse increased the
risk of death from cardiovascular causes (table 4). Hard-drug
use disorder did the same, but not when adjusted for other
substance use disorders. Cannabis use disorder was not
associated with death from cardiovascular causes. In
depression, alcohol use disorder was associated with death
from cardiovascular causes and cannabis was not; hard-drug
use disorders were not associated with cardiovascular deaths
when adjusted for alcohol and cannabis use disorders.
201 (4%) deaths in people with schizophrenia and
164 (4%) deaths in people with depression were due to
respiratory causes (mainly pulmonary cancers, chronic
obstructive pulmonary disease, and pneumonia). All types
of substance use disorders increased the risk of dying from
respiratory causes in people with schizophrenia, although
the association with cannabis did not reach statistical
signifi cance in the fully adjusted model. In depression,
substance use disorders of cannabis, alcohol, and hard
drugs were almost equally signifi cantly associated with
respiratory-related death (table 4).
Schizophrenia (n=41 470) Bipolar disorder (n=11 739) Depression (n=88 269)
All-cause mortality 4616 (11·1%) 856 (7·3%) 3944 (4·5%)
No substance use disorder 1 1 1
Alcohol only 1·52 (1·40–1·65), p<0·0001 1·52 (1·27–1·81), p<0·0001 2·01 (1·86–2·18), p<0·0001
Cannabis only 1·24 (1·04–1·48), p=0·0174 1·09 (0·63–1·86), p=0·7659 0·86 (0·59–1·26), p=0·4414
Hard drugs only 1·78 (1·56–2·04), p<0·0001 1·89 (1·34–2·66), p=0·0003 2·27 (1·98–2·60), p<0·0001
Alcohol and cannabis 1·64 (1·39–1·94), p<0·0001 1·63 (1·09–2·43), p=0·0180 2·36 (1·87–2·98), p<0·0001
Alcohol and hard drugs 2·65 (2·43–2·90), p<0·0001 2·87 (2·35–3·51), p<0·0001 3·41 (3·11–3·75), p<0·0001
Hard drugs and cannabis 2·07 (1·76–2·43), p<0·0001 1·81 (1·06–3·11), p=0·0310 2·71 (2·04–3·60), p<0·0001
Alcohol, cannabis, and hard drugs 2·20 (1·99–2·43), p<0·0001 3·03 (2·34–3·91), p<0·0001 3·44 (2·99–3·96), p<0·0001
Suicide 1138 (2·7%) 251 (2·1%) 998 (1·1%)
No substance use disorder 1 1 1
Alcohol only 0·81 (0·69–0·96), p=0·0141 0·97 (0·70–1·34), p=0·8493 1·04 (0·88–1·22), p=0·6723
Cannabis only 1·28 (0·97–1·76), p=0·0785 0·81 (0·33–1·96), p=0·6330 0·14 (0·03–0·55), p=0·0050
Hard drugs only 1·00 (0·75–1·34), p=0·9798 0·88 (0·41–1·90), p=0·7525 0·97 (0·69–1·34), p=0·8335
Alcohol and cannabis 0·94 (0·67–1·36), p=0·7379 1·02 (0·47–2·21), p=0·9601 1·01 (0·59–1·72), p=0·9826
Alcohol and hard drugs 1·01 (0·81–1·25), p=0·9604 1·04 (0·65–1·68), p=0·8575 1·35 (1·09–1·67), p=0·0061
Hard drugs and cannabis 1·19 (0·85–1·66), p=0·3076 1·75 (0·74–4·15), p=0·2023 0·69 (0·31–1·56), p=0·3751
Alcohol, cannabis, and hard drugs 0·65 (0·49–0·86), p=0·0021 1·06 (0·57–1·99), p=0·8527 0·85 (0·56–1·27), p=0·4277
Deaths from accidents 881 (2·1%) 119 (1·0%) 562 (0·6%)
No substance use disorder 1 1 1
Alcohol only 1·62 (1·28–2·06), p<0·0001 2·06 (1·20–3·52), p=0·0084 2·92 (2·24–3·80), p<0·0001
Cannabis only 1·59 (1·02–2·47), p=0·0398 1·41 (0·33–6·00), p=0·6380 1·80 (0·79–4·10), p=0·1593
Hard drugs only 4·86 (3·71–6·38), p<0·0001 3·06 (1·28–7·32), p=0·0117 6·16 (4·43–8·57), p<0·0001
Alcohol and cannabis 2·40 (1·61–3·56), p<0·0001 1·31 (0·31–5·56), p=0·7099 6·01 (3·62–9·97), p<0·0001
Alcohol and hard drugs 6·74 (5·50–8·25), p<0·0001 5·22 (3·03–9·01), p<0·0001 9·22 (7·12–11·92), p<0·0001
Hard drugs and cannabis 5·78 (4·33–7·72), p<0·0001 2·07 (0·49–8·79), p=0·3257 10·63 (6·71–16·84), p<0·0001
Alcohol, cannabis, and hard drugs 5·46 (4·40–6·76), p<0·0001 9·41 (5·34–16·58), p<0·0001 11·85 (8·71–16·13), p<0·0001
Data are n (%); HR (95% CI), p value; or SHR (95% CI), p value. HRs are presented for all-cause mortality and SHRs for cause-specifi c mortality. All models were adjusted for age
at onset of mental illness, birth year, calendar year, education, and sex. Exposure categories are lifetime substance use disorders. HR=hazard ratio. SHR=subhazard ratio.
Table 3: Substance use disorders as predictors of all-cause mortality, suicide, and deaths from accidents in people with schizophrenia, bipolar disorder,
or depression
Articles
6
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
226 (5%) deaths in people with schizophrenia and
250 (6%) deaths in people with depression were due to
causes in the digestive organs. Alcohol use disorder
signifi cantly increased the risk of death from causes related
to the digestive organs, both in schizophrenia and
depression. Cannabis use disorder was strongly associated
with a decreased risk of dying from diseases in the digestive
organs, albeit only signifi cantly in schizophrenia (table 4).
In sensitivity analyses treating substance use disorders as
time-varying covariates, we noted generally increased
associations between mortality and alcohol and hard-drug
use disorders, but not cannabis use disorder (appendix).
The fi ndings of this sensitivity analysis imply that use
disorders of alcohol and hard drugs were associated with
an increased risk of suicide (appendix). In further sensitivity
analyses, we used substance-related information from only
the hospital-based registers. This had little eff ect on results,
except that cannabis use disorder increased risk of death
from accidents in people with depression (SHR 2·51,
95% CI 1·28–4·92; p=0·0073). Adjusting for seasonal
variations did not change our results appreciably (data not
shown). The associations between alcohol and cannabis
use disorders with suicide in people with schizophrenia
were driven by suicide death through violent methods (data
not shown). Adjusting for multiple psychiatric hospital
admissions had only a negligible eff ect on results, except
that alcohol use disorder became associated with a
decreased risk of suicide in people with depression
(SHR 0·80, 95% CI 0·68–0·95; p=0·0090). Interaction
analyses for sex were either not signifi cant or highly similar
to analyses without interaction terms (data not shown). The
one exception to this was that alcohol use disorder in
women (but not in men) increased risk of suicide in
depression (SHR 1·35, 95% CI 1·04–1·74; p=0·0228).
Discussion
We found that all three types of substance use disorders
were linked to excess mortality in people with schizo-
phrenia, particularly alcohol and hard drugs. For bipolar
disorder and depression, alcohol and hard drugs were also
associated with increased mortality, whereas cannabis was
not. Although not implying causality, SMRs suggested
that a diagnosed substance-use disorder accounted for a
large proportion of the increased mortality in severe
mental illness.
Deaths from accidents were prevalent, and were mostly
accidental poisonings (including substance overdoses) or
poisonings with undetermined intent (data not shown).
The increased risk of death in people with substance use
disorders might be attributable to poisonings related to
the misused substances themselves, to increased risk-
taking behaviour, or to decreased help-seeking abilities.
Alcohol use disorder was linked to a reduced risk of
suicide in people with schizophrenia, and cannabis use
disorder was linked to a reduced risk of suicide in those
with depression. These results are surprising because
alcohol is used in many suicide attempts.22,23 One
explanation could be that psychiatric patients might use
Schizophrenia (n=41 470) Depression (n=88 269)
Model 1: Each type of abuse
entered individually,
unadjusted for one another
Model 2: All types of abuse
entered simultaneously, adjusted
for one another
Model 1: Each type of abuse
entered individually,
unadjusted for one another
Model 2: All types of abuse
entered simultaneously, adjusted
for one another
Malignant tumour-related mortality (except in respiratory organs)
Cannabis use disorder 0·74 (0·47–1·16), p=0·1879 0·70 (0·42–1·15), p=0·1543 0·87 (0·47–1·60), p=0·6490 0·67 (0·36–1·27), p=0·2241
Alcohol use disorder 1·09 (0·83–1·43), p=0·5585 1·12 (0·83–1·50), p=0·4623 0·92 (0·71–1·19), p=0·5279 0·78 (0·59–1·03), p=0·0859
Hard-drug use disorder 1·00 (0·72–1·40), p=0·9995 1·06 (0·72–1·55), p=0·7816 1·78 (1·33–2·37), p<0·0001 2·03 (1·48–2·78), p<0·0001
Cardiovascular-related mortality
Cannabis use disorder 1·06 (0·74–1·51), p=0·7617 0·83 (0·56–1·24), p=0·3712 1·05 (0·57–1·93), p=0·8731 0·74 (0·39–1·42), p=0·3701
Alcohol use disorder 1·94 (1·47–2·56), p<0·0001 1·90 (1·41–2·57), p<0·0001 2·15 (1·52–3·04), p<0·0001 2·01 (1·38–2·91), p=0·0002
Hard-drug use disorder 1·39 (1·04–1·85), p=0·0241 1·16 (0·83–1·62), p=0·3848 1·70 (1·18–2·46), p=0·0044 1·44 (0·95–2·18), p=0·0858
Respiratory organ-related mortality
Cannabis use disorder 2·06 (1·47–2·90), p<0·0001 1·42 (0·98–2·07), p=0·0658 3·21 (2·03–5·09), p<0·0001 2·06 (1·26–3·38), p=0·0041
Alcohol use disorder 2·08 (1·55–2·79), p<0·0001 1·67 (1·20–2·31), p=0·0021 2·17 (1·57–2·99), p<0·0001 1·71 (1·21–2·42), p=0·0023
Hard-drug use disorder 2·23 (1·67–2·98), p<0·0001 1·63 (1·16–2·29), p=0·0053 2·55 (1·83–3·56), p<0·0001 1·85 (1·27–2·69), p=0·0013
Digestive organ-related mortality
Cannabis use disorder 0·68 (0·44–1·03), p=0·0679 0·44 (0·28–0·68), p=0·0002 1·13 (0·67–1·90), p=0·6468 0·60 (0·35–1·03), p=0·0652
Alcohol use disorder 6·62 (4·65–9·42), p<0·0001 7·05 (4·90–10·14), p<0·0001 35·29 (19·29–64·56), p<0·0001 34·02 (18·56–62·33), p<0·0001
Hard-drug use disorder 1·59 (1·19–2·12), p=0·0018 1·11 (0·81–1·50), p=0·5197 2·39 (1·82–3·14), p<0·0001 1·31 (0·99–1·73), p=0·0579
Data are subhazard ratios (95% CIs) and p values versus people without that particular type of substance use disorder. Of people with schizophrenia, 215 (0·5%) died from malignant tumour, 220 (0·5%) died as a
result of cardiovascular problems, 201 (0·5%) died from respiratory organ disorders, and 226 (0·5%) died from digestive organ disorders. Of people with depression, 264 (0·3%) died from malignant tumour,
152 (0·2%) died as a result of cardiovascular problems, 164 (0·2%) died from respiratory organ disorders, and 250 (0·3%) died from digestive organ disorders. All models were adjusted for age at onset of
depression, birth year, calendar year, education, and sex. Exposure categories are lifetime substance use disorders. We did not analyse data for people with bipolar disorder due to insuffi cient number of deaths.
Table 4: Substance use disorders as predictors of natural causes of death in people with schizophrenia or depression
Articles
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
7
substances to self-medicate. This substance use might
reduce symptoms suffi ciently to reduce suicidal ideation,
or these people might cope with suicidal ideation by
using substances. Another explanation could be that
people who use substances are less prone to suicide—eg,
through increased social activity.24 If substance-related
suicides are misclassifi ed in the registers as, for example,
accidental poisonings, this would also explain this
unexpected fi nding. Furthermore, at a p value of 0·01,
this fi nding would not stand correction for multiple
comparisons. Finally, Schoenfeld residuals indicated
deviations from the assumption of proportional hazards
for the alcohol–suicide link in schizophrenia.
Cannabis use disorder was associated with increased
risk of mortality in schizophrenia, but not in bipolar
disorder or depression. In psychosis, cannabis has been
linked to poorer adherence to medication and more severe
psychotic symptoms.25 The increased mortality of this
population could be mediated through such factors. People
with schizophrenia might also use cannabis in higher
quantities than those with bipolar disorder or depression.
People with schizophrenia might be more susceptible to
the eff ects of cannabis than those with bipolar disorder or
depression—eg, as a result of poorer social networks and
increased levels of tobacco use.26 Finally, with a p value of
0·02, this association could be spurious.
Alcohol use disorder, but not hard-drug or cannabis
use disorder, was linked to increased risks of dying from
cardiovascular causes. This replicates consistent previous
fi ndings for alcohol, but previous studies have also linked
both opioids and cocaine to cardiovascular mortality.27–29
Surprisingly, alcohol and cannabis use disorders were
not associated with an increased risk of dying from
malignant tumours, which is usually a consistent
fi nding.30 An increased risk of dying from malignant
tumours has also previously been reported for people
with depression and use of hard drugs.31 We did not fi nd
the same association in schizophrenia, perhaps because
this population was younger than the depression and
bipolar disorder populations. Another surprising fi nding
was that people with cannabis use disorder were at
decreased risk of dying from causes in the digestive
organs; future studies could explore this further.
Tobacco smoking has been identifi ed as a leading
underlying cause of death in people with severe mental
illness.32,33 The association between cannabis and mortality
from causes related to respiratory organs might be
attributable to unmeasured confounding from tobacco
use, although information about tobacco use was not
available. However, around half of people with
schizophrenia and a third of people with bipolar disorder
also use tobacco,34,35 leaving little room left for unmeasured
confounding. Finally, if tobacco use was an important
unmeasured confounder, signifi cant associations between
cannabis use and deaths from cardiovascular causes and
tumours would also be expected, but we did not fi nd any
such associations.
We followed up the entire population of individuals born
in Denmark with relevant diagnoses over an extended
period of time and did mutual adjustment for the diff erent
types of substance use. Furthermore, we maximised
information about substance use by combining infor-
mation from several registers. We chose to investigate
substance-related mortality exclusively in people with
severe mental illness, since substance use disorders might
be less under-diagnosed in this group than in the
general population. Furthermore, the association between
substance use and death from suicide or accidental deaths
might be specifi c to people with severe mental illness.
However, our study has several limitations. Substance
use disorders are under-diagnosed, even in people with
mental illness,36 which would result in an underestimation
of the associations. Furthermore, by limiting our analyses
to individuals with a diagnosed substance use disorder, we
have not assessed the mortality eff ects of drinking or
drug taking that do not meet the thresholds of substance
use disorder. Additionally, in the primary analyses, we
deliberately introduced an immortality bias; individuals in
the exposed categories (ie, those who were diagnosed with
substance use disorder) by defi nition would have survived
to the point of diagnosis, whereas individuals without
diagnosed substance use disorder could have died at any
point during follow up. Therefore, the mortality rate in
substance users will be under-estimated because they were
observed for a longer period than individuals without a
diagnosis of substance use disorder, and could not have
died during the period between diagnosis of mental
disorder and diagnosis of substance use disorder. The HRs
are consequently conservatively biased. Furthermore,
SMRs do not account appropriately for diff erences in
socio economic position. However, results from Cox
regression analyses that adjusted for socioeconomic
position revealed the same increased mortality (data not
shown). An individual could be part of more than one
study population, but adjustment for comorbid psychiatric
disorders did not aff ect results (data not shown). We
classifi ed people who had received prescription medicines
used to treat alcohol and hard-drug use disorders as having
an alcohol or hard-drug use disorder; however, some
people could have received these medications for severe
somatic illness. Nonetheless, sensitivity analyses without
prescription medicine data had little if any eff ect on the
results (data not shown). We did not correct for multiple
comparisons, because this would substantially increase
the rate of type II errors. Consequently, p values around
0·05 are interpreted cautiously. Interaction analyses for
sex were highly similar to analyses without interaction
terms, except that alcohol use disorder in women (but not
in men) increased risk of suicide in depression. Plots of
Schoenfeld residuals generally confi rmed the assumptions
of proportional hazards. The one exception was the
apparent reduced risk of suicide in people with alcohol use
disorders in the schizophrenia population, which is why
this particular result should be interpreted cautiously.
Articles
8
www.thelancet.com/psychiatry Published online August 13, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00207-2
Our fi ndings raise the hypothesis that successful
treatment of substance use disorders might reduce
mortality in people with severe mental illness. Un-
fortunately, reviews have shown that treating substance
use disorders in people with severe mental illness is often
not eff ective.37,38 Nonetheless, awareness of substance
misuse in people with mental illness could help clinical
eff orts to reduce excess mortality. Even in the absence of
causality, patients with dual diagnosis are still an important
target group for mortality-reducing interventions. Our
fi ndings also highlight the importance of preventing
substance use disorders. In conclusion, neither alcohol,
hard drugs, nor cannabis should be considered harmless.
Contributors
CH and JTA designed the study protocol. CH obtained funding, did the
scientifi c literature search, and obtained and analysed the data. CH,
MLDØ, and NGT developed the algorithm for substance use disorders.
MEB, MLDØ, AE, JTA, and MN critically revised the data analysis plan
initially proposed by CH. All authors interpreted the data. CH wrote the
fi rst draft of the manuscript, and all authors critically revised this and
approved the fi nal version of the manuscript.
Declaration of interests
We declare no competing interests.
Acknowledgments
We thank the Lundbeck Foundation who funded the study.
References
1 Walker ER, McGee RE, Druss BG. Mortality in mental disorders
and global disease burden implications: a systematic review and
meta-analysis. JAMA Psychiatry 2015; 72: 334–41.
2 Laursen TM, Munk-Olsen T, Nordentoft M, Mortensen PB.
Increased mortality among patients admitted with major psychiatric
disorders: a register-based study comparing mortality in unipolar
depressive disorder, bipolar aff ective disorder, schizoaff ective
disorder, and schizophrenia. J Clin Psychiatry 2007; 68: 899–907.
3 Nordentoft M, Wahlbeck K, Hallgren J et al. Excess mortality, causes
of death and life expectancy in 270 770 patients with recent onset of
mental disorders in Denmark, Finland and Sweden. PLoS One 2013;
8: e55176.
4 Hiroeh U, Kapur N, Webb R, Dunn G, Mortensen PB, Appleby L.
Deaths from natural causes in people with mental illness: a cohort
study. J Psychosom Res 2008; 64: 275–83.
5 Hiroeh U, Appleby L, Mortensen PB, Dunn G. Death by homicide,
suicide, and other unnatural causes in people with mental illness:
a population-based study. Lancet 2001; 358: 2110–12.
6 Harris EC, Barraclough B. Excess mortality of mental disorder.
Br J Psychiatry 1998; 173: 11–53.
7 Green B, Young R, Kavanagh D. Cannabis use and misuse prevalence
among people with psychosis. Br J Psychiatry 2005; 187: 306–13.
8 Koskinen J, Lohonen J, Koponen H, Isohanni M, Miettunen J.
Prevalence of alcohol use disorders in schizophrenia—a systematic
review and meta-analysis. Acta Psychiatr Scand 2009; 120: 85–96.
9 Koskinen J, Lohonen J, Koponen H, Isohanni M, Miettunen J.
Rate of cannabis use disorders in clinical samples of patients with
schizophrenia: a meta-analysis. Schizophr Bull 2010; 36: 1115–30.
10 Regier DA, Farmer ME, Rae DS, et al. Comorbidity of mental
disorders with alcohol and other drug abuse. Results from the
Epidemiologic Catchment Area (ECA) study. JAMA 1990; 264: 2511–18.
11 Koola MM, McMahon RP, Wehring HJ, et al. Alcohol and cannabis
use and mortality in people with schizophrenia and related
psychotic disorders. J Psychiatr Res 2012; 46: 987–93.
12 Aldington S, Harwood M, Cox B, et al. Cannabis use and risk of
lung cancer: a case-control study. Eur Respir J 2008; 31: 280–86.
13 Bachs L, Morland H. Acute cardiovascular fatalities following
cannabis use. Forensic Sci Int 2001; 124: 200–03.
14 Pedersen CB. The Danish Civil Registration System.
Scand J Public Health 2011; 39 (suppl): 22–25.
15 Mors O, Perto GP, Mortensen PB. The Danish Psychiatric Central
Research Register. Scand J Public Health 2011; 39 (suppl): 54–57.
16 Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient
Register. Scand J Public Health 2011; 39 (suppl): 30–33.
17 Kildemoes HW, Sorensen HT, Hallas J. The Danish National
Prescription Registry. Scand J Public Health 2011; 39 (suppl): 38–41.
18 Danmarks Statistik. Register over stofmisbrugere i behandling
(SIB). http://www.ssi.dk/Sundhedsdataogit/Registre%20og%20
kliniske%20databaser/De%20nationale%20sundhedsregistre/
Sygdomme%20leagemidler%20behandlinger/Stofmisbrugere%20
i%20behandling.aspx (accessed July 22, 2015).
19 Helweg-Larsen K. The Danish Register of Causes of Death.
Scand J Public Health 2011; 39 (suppl): 26–29.
20 Petersson F, Baadsgaard M, Thygesen LC. Danish registers on
personal labour market affi liation. Scand J Public Health 2011;
39 (suppl): 95–98.
21 Fine JP, Gray RJ. A proportional hazards model for the subdistribution
of a competing risk. J Am Stat Assoc 1999; 94: 496–509.
22 Wilcox HC, Conner KR, Caine ED. Association of alcohol and drug
use disorders and completed suicide: an empirical review of cohort
studies. Drug Alcohol Depend 2004; 76 (suppl): S11–19.
23 Larkin C, Di BZ, Arensman E. Risk factors for repetition of
self-harm: a systematic review of prospective hospital-based studies.
PLoS One 2014; 9: e84282.
24 Drake RE, Wallach MA, Alverson HS, Mueser KT. Psychosocial
aspects of substance abuse by clients with severe mental illness.
J Nerv Ment Dis 2002; 190: 100–06.
25 Zammit S, Moore THM, Lingford-Hughes A, et al. Eff ects of
cannabis use on outcomes of psychotic disorders: systematic review.
Br J Psychiatry 2008; 193: 357–63.
26 Calabria B, Degenhardt L, Hall W, Lynskey M. Does cannabis use
increase the risk of death? Systematic review of epidemiological
evidence on adverse eff ects of cannabis use. Drug Alcohol Rev 2010;
29: 318–30.
27 Ronksley PE, Brien SE, Turner BJ, Mukamal KJ, Ghali WA.
Association of alcohol consumption with selected cardiovascular
disease outcomes: a systematic review and meta-analysis. BMJ 2011;
342: d671.
28 Degenhardt L, Bucello C, Mathers B, et al. Mortality among regular
or dependent users of heroin and other opioids: a systematic review
and meta-analysis of cohort studies. Addiction 2011; 106: 32–51.
29 Phillips K, Luk A, Soor GS, Abraham JR, Leong S, Butany J.
Cocaine cardiotoxicity: a review of the pathophysiology, pathology,
and treatment options. Am J Cardiovasc Drugs 2009; 9: 177–96.
30 Boff etta P, Hashibe M, La Vecchia C, Zatonski W, Rehm J. The
burden of cancer attributable to alcohol drinking. Int J Cancer 2006;
119: 884–87.
31 Randall D, Degenhardt L, Vajdic CM, et al. Increasing cancer
mortality among opioid-dependent persons in Australia: a new
public health challenge for a disadvantaged population.
Aust NZ J Public Health 2011; 35: 220–25.
32 Kelly DL, McMahon RP, Wehring HJ, et al. Cigarette smoking and
mortality risk in people with schizophrenia. Schizophr Bull 2011;
37: 832–38.
33 Suvisaari J, Partti K, Perala J, et al. Mortality and its determinants in
people with psychotic disorder. Psychosom Med 2013; 75: 60–67.
34 Poirier MF, Canceil O, Bayle F et al. Prevalence of smoking in
psychiatric patients. Prog Neuropsychopharmacol Biol Psychiatry 2002;
26: 529–37.
35 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.
36 Hansen SS, Munk-Jørgensen P, Guldbaek B, et al. Psychoactive
substance use diagnoses among psychiatric in-patients.
Acta Psychiatr Scand 2000; 102: 432–38.
37 Hunt GE, Siegfried N, Morley K, Sitharthan T, Cleary M. Psychosocial
interventions for people with both severe mental illness and
substance misuse. Cochrane Database Syst Rev 2013; 10: CD001088.
38 Hjorthøj C, Fohlmann A, Nordentoft M. Treatment of cannabis use
disorders in people with schizophrenia spectrum disorders—a
systematic review. Addict Behav 2009; 34: 520–25.