Original Contribution

Childhood Social Environment and Risk of Drug and Alcohol Abuse in a Cohort
of Danish Men Born in 1953
Merete Osler
1,2
, Merete Nordentoft
3
, and Anne-Marie Nybo Andersen
4
1
Department of Social Medicine, Institute of Public Health, University of Copenhagen, Copenhagen, Denmark.
2
Epidemiology Unit, Institute of Public Health, University of Southern Denmark, Odense, Denmark.
3
Department of Psychiatry, Bispebjerg Hospital, Copenhagen, Denmark.
4
National Institute of Public Health, Copenhagen, Denmark.
Received for publication May 30, 2005; accepted for publication October 28, 2005.
In a 32-year follow-up study, the authors analyze how social circumstances during early life, childhood social
participation, and school performance affect the risk of being admitted to a hospital or dying from a diagnosis
closely related to drug or alcohol abuse in young adulthood. A total of 11,376 Danish males born in 1953, for whom
data from birth certicates and conscription board examinations had been traced, were followed until 2002 through
linkage to the Danish Psychiatric, National Patient, and Cause of Death registries. At age 12 years, 7,877 subjects
completed a questionnaire on social participation and school performance. During follow-up, 12 percent of these
were given a diagnosis indicating drug or alcohol abuse. Having a single mother and a working-class father were
each associated with an increased risk of drug or alcohol abuse in adult life. At age 12 years, those who disliked
school, scored low on a school test, or preferred to visit a youth club during leisure time showed a greater risk of
adult substance abuse. These associations were slightly attenuated when adjusted for educational status at
conscription. Deprived social circumstances during childhood, poor school performance in early adolescence,
and attending a youth club seemed to be independent markers of substance abuse in adult life.
alcoholism; cohort studies; social environment; substance-related disorders
Abbreviation: ICD, International Classication of Diseases.
There are considerable economic, health, and social costs
associated with drug and alcohol abuse (1), and extensive
research has been devoted to identifying factors associated
with substance use and abuse (111). Prospective studies
that explore early determinants of substance abuse later in
adult life, however, are scarce. A few studies have shown
a relation between socioeconomic disadvantages during
childhood and increased risk of death due to substance
abuse (1215). Several nonexclusive explanations for this
association exist. One such explanation could be the con-
ditions of the family environment during early life. Poor
material conditions and social support, which tend to cluster
in families primarily for social reasons, would inuence the
behavior and health of the members of these families (11,
12). Several investigations, mostly cross-sectional studies,
have shown that poor socioeconomic circumstances, paren-
tal substance abuse, parental divorce, negative attitudes to-
wards school, family poverty, and peer relationships during
childhood are associated with substance use in adolescence.
The effects of these factors can either act independently of
each other or be mediated through the childs own social
position later in life.
In the present study, we examine the effect of parental
socioeconomic position, childhood social participation, sat-
isfaction with school, and school performance on later con-
tinuous and problematic substance use, which is indicated
Correspondence to Dr. Merete Osler, Epidemiology Unit, Institute of Public Health, University of Southern Denmark, J. B. Winslws Vej 9B,
5000 Odense C, Denmark (e-mail: mosler@health.sdu.dk).
654 Am J Epidemiol 2006;163:654661
American Journal of Epidemiology
Copyright 2006 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved; printed in U.S.A.
Vol. 163, No. 7
DOI: 10.1093/aje/kwj084
Advance Access publication January 27, 2006

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

by the risk of being admitted to a hospital or dying from
a diagnosis related to long-term drug or alcohol abuse in
adult life. We also explore whether any such effects are
independent of the subjects own education later in life.
MATERIALS AND METHODS
Study population
According to ofcial statistics, 12,270 boys were born
within the metropolitan area of Copenhagen in 1953. These
persons formed the population of the Danish longitudinal
study (Project Metropolit), which has been described in detail
elsewhere (16). Briey, 11,376 members of this population
who were alive and living in Denmark in 1968 were regis-
tered with a unique, personal identication number (referred
to as a CPR number) when the Civil Registration System
was established. In 1965, a total of 7,877 (69.2 percent) of
these males participated in a school-based survey.
Data sources
Data from birth certicates, including information on the
date and place of birth, mothers marital status, and fathers
occupational status at the time of delivery, were collected
manually for all the members of the original study popula-
tion. The school-based survey, in 1965, included a question-
naire administered by the subjects teachers. Each pupil
provided his name, birth date, place of birth, and fathers
occupation. The survey involved tests of cognition and ques-
tions regarding social aspirations and leisure-time activities.
As part of the conscription procedure, all Danish men un-
dergo physical and mental examinations when they are
about 18 years of age (17). In 2004, we collected data from
the conscript registers for all cohort members who were
alive in 1971 (n 11,337).
Variables
Parental characteristics. Mothers marital status at the
time of delivery was treated in three categories: married;
unmarried (i.e., single, divorced, widowed); and unknown.
Fathers occupation at the time of birth was recorded in
23 strata (nonurban self-employed (four strata); urban self-
employed (six strata); salaried employed (ve strata);
skilled workers and unskilled workers (ve strata); pen-
sioners; students; and unknown). In the preliminary data
analyses, we computed risk estimates for various combina-
tions of the 23 strata. This exercise suggested that they could
be combined into three categories (high/middle, which in-
cluded the self-employed and salaried employed; working
class, which included skilled and unskilled workers; and
unknown, which included the groups of pensioners (n 1),
students (n 44), and fathers not known).
School and leisure-time characteristics. Social partici-
pation at age 12 years was measured by four among 33
different items from the school survey identifying preferred
leisure-time activity. These items were the following: 1) to
be at home with friends; 2) to be at home with family; 3) to
meet with friends; or 4) to visit a youth club. School sat-
isfaction at age 12 years was measured by the question of
whether or not the boy liked to go to school. Cognitive
performance was measured by the Harnquist school test
(14). This test consisted of spatial, arithmetic, and verbal
subtests. Each subtest contained 40 problems, with one
point awarded for each correct answer.
Conscript characteristics. Educational level, primarily
reecting school education, was registered on a scale rang-
ing from leaving school after the eighth grade to attaining
the approximate equivalent of the British advanced-level
general Certicate of Education. In the present analysis, this
information was recoded into three categories.
Follow-up
In August 2004, the Metropolit cohort was followed up for
vital status through record linkage with the Civil Registration
System registry. If a subject was not alive or living outside
Denmark, the date of death or the date of emigration/disap-
pearance was obtained. Information on the time of admission
to psychiatric wards from 1969 to January 2003, as well as
diagnosis on discharge, was obtained from the Danish Psy-
chiatric Central Register. This register has compiled comput-
erized data on admissions to psychiatric hospitals and to
psychiatric departments in general hospitals, in Denmark,
since April 1969 (18). The National Patient Registry pro-
vided information on admission to somatic wards since
1978, when this registry was established (19). Causes of
death from1968 to December 2001 were determined through
record linkage with the Cause of Death Registry (20).
Diagnoses were classied according to the International
Classication of Diseases (ICD), Eighth Revision, for the
years 19691993 and according to the ICD, Tenth Revision,
from 1994 onward. Drug and alcohol abuse were dened
according to the ICD codes in table 1.
TABLE 1. Diagnoses and ICD* codes used for outcome
denitions in a study of 7,877 Danish men born in 1953
Diagnosis ICD-8* code ICD-10* code
Drug related
Opioids 304.09, 304.19 F11.0F11.9
Cannabinoids 304.59 F12.0F12.9
Sedatives/hypnotics 304.29, 304.39 F13.0F13.9
Cocaine 304.49 F14.0F14.9
Other stimulants 304.69 F15.015.9
Hallucinogens 304.79 F16.0F16.9
Other and multiple drugs 304.89, 304.99 F18.0F19.9
Alcohol related
Alcohol psychosis and
abuse syndrome 291.09291.99 F10.0F10.9
303.09303.99 K70.0K70.9
Cirrhosis of the liver 571.09, 571.19 I85.0I85.9
Esophageal varices 456.09, 456.19
* ICD, International Classication of Diseases; ICD-8, ICD, Eighth
Revision; ICD-10, ICD, Tenth Revision.
Social Environment and Substance-related Disorders 655
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Statistical methods
Associations between social variables in early life and
drug or alcohol abuse were analyzed using Coxs propor-
tional hazards regression model with age as the underlying
time scale. The events were dened as death from or rst-
time discharge from a hospital with a substance abuse-
related diagnosis, as described in table 1. Entry time was
the subjects age on April 1, 1969, and follow-up ended at
the time of event, time of emigration from Denmark, death
from nonevents, or January 1, 2003, whichever came rst.
Since follow-up in the National Patient Registry started in
1978 and most conscription board examinations took place
in 1971 and 1972, we repeated all analyses with the age in
1978 as the entry. These two approaches, however, gave
essentially the same results. In the present paper, therefore,
we report only the rst. The proportional hazards assump-
tion was evaluated for all variables by comparing estimated
-lognormal(-lognormal) survivor curves over the different
categories of variables being investigated versus lognormal
(analysis time) and by tests based on the generalization as
described by Grambsch and Therneau (21). Statistical anal-
yses were computed using STATA, version 7, software
(Stata Corporation, College Station, Texas).
RESULTS
By follow-up, a total of 316 (4.0 percent) of the 7,877
men who participated in the school survey had died or been
discharged from a hospital with a diagnosis related to drug
abuse (table 2). The median age for rst diagnosis was 25.6
years. For alcohol abuse, the respective values were 696 (8.8
percent) and 28.2 years, respectively. The risk of substance
abuse was slightly higher among men who did not attend the
school survey.
The distributions of fathers occupational class and moth-
ers marital status at the subjects birth are shown in table 3,
along with the unadjusted and mutually adjusted hazard
ratios for drug and alcohol abuse according to these two
characteristics. Both indicators of poor socioeconomic cir-
cumstances during early life were associated with an in-
creased risk of discharge with a drug abuse diagnosis
before and after mutual adjustments (table 3). A similar
pattern of associations was seen for alcohol abuse. Crude
analyses for those of the 3,499 nonparticipants for whom
data were available on the mothers marital status and fa-
thers occupational class at the subjects birth produced risk
estimates in the same direction as those based on the 7,877
subjects who participated in the school survey.
Compared with those whose parents were of higher social
positions, men who were born to single mothers or had
working-class fathers were more likely to select youth-club
visits as their favorite leisure-time activity, to dislike school,
and to have lower scores on the school test at age 12 years.
Preferring to be at home, with either friends or family, dur-
ing leisure time was not associated with paternal occupa-
tional class or mothers marital status (data not shown).
Those with youth-club visits as their leisure-time prefer-
ence had a signicantly increased risk of drug and alcohol
abuse when adjustments were made for other school and
leisure-time characteristics (table 4). Further, those who dis-
liked school or scored low on the school test at age 12 years
had an increased risk of substance abuse. The effects of
indicators of social circumstances during early life were
only slightly attenuated by the inclusion of indicators of
social participation, satisfaction with school, and perfor-
mance at age 12 years in the multivariate model (table 5).
The effect of the latter characteristics did not vary in relation
to social indicators (no signicant interactions).
The lowest educational level at the conscription board
examination was most often found among men who were
born to single mothers, had working-class fathers, preferred
to visit youth clubs, did not like school, or scored low on the
school test at age 12 years. Educational status at that exam-
ination was also inversely associated with the risk of both
drug and alcohol abuse, and the risk estimates for social
circumstances during early life, childhood social participa-
tion, and school performance changed slightly when adult
educational level was included in the model (table 6).
DISCUSSION
In this cohort of Danish men born in 1953 who grew up
during a youth rebellion that featured rock music, drugs, and
sexual liberation, we found that the risk of being admitted to
a hospital or dying from a condition related to drug and
alcohol abuse was surprisingly high. Based on rather strin-
gent criteria of abuse, namely, death or hospital admission
and diagnoses associated with long-term use, the risk was 12
TABLE 2. Number of cases of drug and alcohol abuse among Danish men born in the metropolitan area of Copenhagen in 1953
Participated in school survey (n 7,877) Did not participate in school survey (n 3,399)
Drug abuse
diagnosis
Alcohol abuse
diagnosis
Drug and/or alcohol
abuse diagnosis
Drug abuse
diagnosis
Alcohol abuse
diagnosis
Drug and/or alcohol
abuse diagnosis
No. % No. % No. % No. % No. % No. %
Psychiatric register 261 3.4 399 5.7 536 6.8 151 4.4 211 6.2 285 8.3
Somatic register 108 1.4 457 5.8 509 6.5 75 2.2 238 7.0 271 4.0
Death register 72 0.9 161 2.0 227 2.9 51 1.5 71 2.0 153 4.5
Total* 316 4.0 696 8.8 946 12.0 190 5.4 352 10.3 442 13.0
* Number of persons with appearance in at least one of the above-mentioned registers.
656 Osler et al.
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

percent. Indicators of poor social circumstances during early
life were associated with an increased risk of adult substance
abuse. Youth-club visits as a preferred leisure-time interest,
poor satisfaction with school, and a low performance test
score at age 12 years also increased the risk of substance
abuse. These effects were slightly attenuated after adjust-
ment for educational status around the age of 18 years,
which itself was inversely related to both drug and alcohol
abuse. We found no effect associated with preferring to
spend leisure time at home with family or friends.
Strengths and limitations
The present data relate to all males who were born in
a well-dened area (covering one third of the Danish pop-
ulation) and who survived to the age of 15 years. We had
prospectively collected information on life conditions and
attitudes at age 12 years and at the conscription board ex-
amination. By using the population-covering registers, we
managed to get complete follow-up information. Thus, hos-
pital admission data were available for more than 95 percent
of this nonselected population. Information on social partic-
ipation, satisfaction with school, and school performance,
however, was available for only 69.2 percent of the cohort
members, and the risk of drug and alcohol abuse was
slightly higher among nonparticipants (13.0 percent vs.
12.0 percent). On the other hand, the risk estimates for the
parental variables that were available for all participants did
not differ much for nonparticipants. The follow-up covered
a period of more than 30 years, but it might take several
years to develop severe complications of alcohol abuse. It
seems likely, therefore, that future follow-up will capture
more cases.
We had information only on disorders diagnosed during
admission to hospitals. Consequently, we assume that our
outcome measure is more sensitive to dropout during fol-
low-up than self-reports, since it does not depend on the
subjects ability to answer a questionnaire. Our operational
denition was rather conservative, because we used diag-
noses closely linked to drug and alcohol abuse and not more
broadly dened alcohol-related outcomes, such as gastric
cancers, injuries, and acute alcohol intoxication. In Den-
mark, admission to hospitals is free, and it is likely that most
of those with a diagnosis related to advanced abuse have
been admitted to a hospital. It does remain possible, how-
ever, that some men affected by drug or alcohol use are
TABLE 3. Risk of drug or alcohol abuse at age 1549 years according to parental characteristics among 7,877 Danish men born
in 1953
No. %
Drug abuse Alcohol abuse
No.
Unadjusted Adjustedy
No.
Unadjusted Adjustedy
Hazard
ratio
95%
condence
interval
Hazard
ratio
95%
condence
interval
Hazard
ratio
95%
condence
interval
Hazard
ratio
95%
condence
interval
Participated in school survey (n 7,877)
Mothers marital
status at birth
Married 7,127 90.5 270 1.00 1.00 596 1.00 1.00
Single 512 6.5 34 1.78 1.24, 2.54* 1.60 1.11, 2.30* 74 1.80 1.41, 2.29* 1.61 1.16, 2.25*
Unknown 238 3.0 12 26
Fathers occupational
status at birth
High/middle 3,504 44.8 105 1.00 1.00 240 1.00 1.00
Working 3,729 47.3 171 1.54 1.31, 1.97 1.53 1.20, 1.58 376 1.50 1.28, 1.77 1.48 1.26, 1.75
Unknown 644 8.2 40 2.10 1.40, 3.04* 1.82 1.02, 3.22* 80 1.89 1.47, 2.43* 1.45 0.98, 2.16*
Did not participate in school survey (n 3,399)
Mothers marital
status at birth
Married 2,918 83.2 150 1.00 1.00 266 1.00 1.00
Single 419 12.0 35 1.64 1.13, 2.38* 1.23 0.76, 1.98 76 2.10 1.63, 2.71* 1.79 1.29, 2.47*
Unknown 168 4.8 5 10
Fathers occupational
status at birth
High/middle 1,488 42.5 53 1.00 1.00 98 1.00 1.00
Working 1,542 44.1 105 1.92 1.38, 2.68 1.90 1.36, 2.63 194 1.96 1.54, 2.51 1.89 1.48, 2.42
Unknown 469 13.4 32 1.99 1.28, 3.09* 2.22 1.23, 4.00* 60 2.10 1.52, 2.84* 1.76 1.15, 2.69*
* p < 0.001 (Walds test).
y The adjusted model includes marital status and occupational class.
Social Environment and Substance-related Disorders 657
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

untreated, treated solely in private specialist practice, or
treated by a general practitioner. Such cases are not included
in this study and might also be missed in studies based on
self-reported substance abuse, because of the presumed low
attendance rate in surveys among heavy abusers.
One might also challenge the contemporary relevance of
the social participation measures we used, since the Copen-
hagen youth club was a social pedagogic experiment de-
signed, in the 1960s, to assist youths in deprived areas. In
addition, it might be that some of the men in this cohort were
just beginning their substance abuse at the time of their
conscription board examinations. We used information on
educational level and not on current occupation, since the
latter is more likely to be a consequence of current interest.
Further, half of the cases had their rst admission at age 25
years or later, and risk estimates were essentially the same
when we used 1978 as the entry time in statistical analyses.
Although substance abuse in cohorts born in the 1950s is
more common in men than in women, it is an obvious lim-
itation to our study that women were not represented in the
data set. In addition, no information was available on some
important determinants such as parental abuse. On the other
hand, previous studies have shown that relations between
adverse childhood experiences and alcohol abuse exist in-
dependently of parental abuse (4, 10).
Comparison with other studies
Few of the increasing number of studies concerning
childhood socioeconomic circumstances and adult mortality
have included outcomes of substance abuse (12). A large,
male cohort in Finland showed more alcohol-related deaths
among men with fathers from a lower social class (13), and,
in the Oslo mortality study, housing conditions during child-
hood were associated with psychiatric deaths due largely to
alcohol or drug dependence (14). A Danish register study,
which included 84,765 children born in 1966, showed that
parental abuse of alcohol was associated with an increased
TABLE 4. Risk of drug or alcohol abuse at age 1549 years according to school and leisure-time characteristics at age 12 years
among 7,877 Danish men born in 1953
Total Drug abuse Alcohol abuse
No. % No.
Unadjusted Adjustedy
No.
Unadjusted Adjustedy
Hazard
ratio
95%
condence
interval
Hazard
ratio
95%
condence
interval
Hazard
ratio
95%
condence
interval
Hazard
ratio
95%
condence
interval
Social participation
at age 12 years
Visits youth club
No 5,651 71.1 185 1.00 1.00 452 1.00 1.00
Yes 2,032 25.8 121 1.84 1.41, 2.31** 1.55 1.22, 1.97** 219 1.36 1.16, 1.60** 1.22 1.04, 1.45*
Meeting with friends
No 3,657 46.4 122 1.00 1.00 289 1.00 1.00
Yes 4,040 51.3 185 1.37 1.08, 1.77* 1.17 0.90, 1.53 386 1.21 1.04, 1.41* 1.13 0.95, 1.34
Being at home
with family
No 2,916 37.0 105 1.00 1.00 233 1.00 1.00
Yes 4,767 60.5 201 1.17 0.92, 1.48 1.07 0.82, 1.38 440 1.15 0.99, 1.35 1.11 0.94, 1.32
Being at home
with friends
No 3,243 41.2 109 1.00 1.00 263 1.00 1.00
Yes 4,439 56.4 197 1.31 1.04, 1.66* 1.10 0.83, 1.45 410 1.14 0.98, 1.33 0.98 0.82, 1.18
Likes to go to school
at age 12 years
Yes 3,916 49.7 126 1.00 1.00 330 1.00 1.00
Somewhat 3,506 44.5 155 1.38 1.09, 1.74 1.41 1.11, 1.79 305 1.03 0.89, 1.28 1.03 0.87, 1.28
No 386 4.9 32 2.63 1.78, 3.88** 2.28 1.52, 3.43** 52 1.64 1.28, 2.20* 1.55 1.15, 2.10*
Cognitive school test
Highest third 2,457 31.2 64 1.00 1.00 148 1.00 1.00
Middle third 2,718 34.5 101 1.43 1.04, 1.96 1.36 0.93, 1.71 219 1.35 1.10, 1.66 1.33 1.08, 1.65
Lowest third 2,590 32.9 147 2.14 1.57, 2.76** 1.64 1.46, 2.67** 317 2.11 1.74, 2.57** 1.81 1.66, 2.47**
* p < 0.05; **p < 0.001.
y The adjusted model includes all the variables in the table.
658 Osler et al.
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

TABLE 5. Risk of drug or alcohol abuse at age 1549 years according to parental,
school, and leisure-time characteristics among 7,877 Danish men born in 1953y
Drug abuse Alcohol abuse
Hazard
ratio
95% condence
interval
Hazard
ratio
95% condence
interval
Mother single at birth 1.37 0.82, 2.28 1.55 1.10, 2.17*
Fathers occupational class
Working vs. high/middle 1.33 1.03, 1.71 1.23 1.13, 1.58**
Unknown vs. high/middle 1.43 0.78, 2.59 1.35 0.91, 2.02
Visits youth club at age 12 years 1.58 1.23, 2.01** 1.21 1.03, 1.44*
Meeting with friends 1.19 0.91, 1.56 1.19 1.00, 1.43
Being at home with family 1.07 0.82, 1.39 1.09 0.92, 1.31
Being at home with friends 1.08 0.82, 1.46 0.97 0.80, 1.18
Likes to go to school at age
12 years
Somewhat vs. yes 1.47 0.95, 1.78 1.02 0.84, 1.20
No vs. yes 2.40 1.59, 3.61** 1.49 1.10, 2.03**
Cognitive school test at age
12 years
Middle vs. highest third 1.38 0.93, 1.78 1.24 1.00, 1.53
Lowest vs. highest third 1.75 1.28, 2.37** 1.81 1.47, 2.22**
* p < 0.05; **p < 0.001.
y The adjusted model includes all the variables in the table.
TABLE 6. Risk of drug or alcohol abuse at age 1549 years according to parental,
school, and leisure-time characteristics among 7,877 Danish men born in 1953y
Drug abuse Alcohol abuse
Hazard
ratio
95% condence
interval
Hazard
ratio
95% condence
interval
Mother single at birth 1.27 0.76, 2.10 1.44 1.03, 2.03*
Fathers occupational class
Working vs. high/middle 1.17 0.90, 1.50 1.11 0.94, 1.32
Unknown vs. high/middle 1.23 0.68, 2.20 1.17 0.78, 1.76
Visits youth club at age 12 years 1.50 1.17, 1.91** 1.15 0.97, 1.37
Meeting with friends 1.23 0.94, 1.61 1.21 1.00, 1.45*
Being at home with family 1.08 0.83, 1.41 1.12 0.93, 1.34
Being home with friends 1.09 0.81, 1.44 0.97 0.81, 1.19
Likes to go to school at age
12 years
Somewhat vs. yes 1.36 1.07, 1.74 0.97 0.82, 1.13
No vs. yes 1.92 1.27, 2.90** 1.28 0.94, 1.74
Cognitive school test at age
12 years
Middle vs. highest third 1.08 0.78, 1.51 0.99 0.74, 1.23
Lowest vs. highest third 1.27 0.91, 1.77 1.20 0.94, 1.48
Educational level at conscript
High 1.00 1.00
Middle 1.90 1.14, 3.35 2.05 1.53, 2.75
Low 5.81 3.41, 9.90 4.72 3.38, 6.44
Did not attend/unknown 11.69 7.00, 19.54** 3.01 2.04, 4.24**
* p < 0.05; **p < 0.001.
y The adjusted model includes all the variables in the table.
Social Environment and Substance-related Disorders 659
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

risk of hospital admission for a diagnosis related to drug
addiction (crude odds ratio 3.3, 95 percent condence
interval: 2.4, 4.5). However, the effect was explained mainly
by factors that were closely linked to parental alcohol abuse
(e.g., long-term parental unemployment, low education, vi-
olence in the family, and mental illness) (4). In the Adverse
Childhood Experience Study (10, 11), a retrospective cohort
study, 9,346 adults who visited a primary care clinic com-
pleted a survey about nine adverse childhood experiences,
including childhood mental and physical abuse, domestic
violence, parental divorce, and growing up with a drug-
abusing or mentally ill household member. Each of the
adverse childhood experiences was associated with self-
reported adult alcoholism, and the number of adverse ex-
periences had a graded relation to alcoholism, independent
of parental alcohol abuse (10). Some cross-sectional and
short-term follow-up studies have analyzed social determi-
nants of adolescent substance use, and a wide range of fac-
tors (e.g., the personality of the young, factors concerning
their family situation, peer-related factors, school and lei-
sure-time activities) have been shown to contribute to the
development of adolescent addictive behavior (2, 9). A few
studies have included factors from more than one category
in order to distinguish relative effects. Those studies that are
available, however, have associated high alcohol consump-
tion and drug use in adolescence with low socioeconomic
position, parental substance abuse, poor family relations,
time spent with peers, peer support and risk behavior, lack
of school adoption, low self esteem, and high alcohol use
in young adulthood (19). The present study suggests that
the family situation, peer factors, and school-related factors
also predict to continuous and problematic substance use, as
indicated by a higher risk of substance abuse requiring med-
ical treatment in adulthood.
Interpretation
The present study indicates that social disadvantages dur-
ing childhood, dissatisfaction with school, and poor school
performance are related to the risk of hospital admission or
death from drug or alcohol abuse in young adult men. To our
surprise, preference for youth-club participation was also
associated with higher risks. This might reect the fact that
1960s Copenhagen youth clubs were attended primarily by
low-income youths. Multiple factors, such as social rela-
tions and psychological stress, have been suggested as me-
diators in the relation between early social circumstances
and substance abuse later in life (11, 12). In our study, the
occupational class variable might reect material wealth,
parental attitudes, and behavior, while the risk associated
with single mothers also reects limited resources in child-
hood. These factors, in turn, might inuence social partici-
pation, school performance, and the behavior of offspring.
Our ndings indicate that prevention of drug and alcohol
abuse in families and schools should start early. Medical and
public health practitioners and teachers should also have an
increased awareness of adverse childhood conditions and
signs of poor performance. Since dissatisfaction with school
and poor school performance in early adolescence seem to
be rather strong predictors of adult substance abuse, parents
and schools should be aware of these early markers and their
social origins. Further, both parents and school teachers
should appreciate the importance of a pleasant and inspiring
school environment. In addition, youth clubs should con-
sider students access to alcohol and drugs in the institution.
The high prevalence of drug and alcohol abuse in these
Danish men also indicates that the liberal Danish drug pol-
icy, especially in relation to alcohol (22), is problematic and
needs to be revised.
ACKNOWLEDGMENTS
Supported by the Danish Heart Association, the Lund-
beck Foundation, and the Danish Health Insurance funds.
The authors thank all those who initiated and/or contin-
ued the Metropolit Study: K. Svalastoga, E. Hgh, P. Wolf,
T. Rishj, G. Strande-Srensen, E. Manniche, B. Holten,
I. A. Weibull, and A. Ortman.
Conict of interest: none declared.
REFERENCES
1. Galea S, Ahern J, Vlahov D. Contextual determinants of drug
use risk behaviour: a theoretical framework. J Urban Health
2003;80(suppl 3):iii508.
2. Hoffmann JP, Su SS. Paternal substance use disorder, medi-
ating variables and adolescent drug use: a non-recursive
model. Addiction 1998;93:135164.
3. Friedman AS, Glassman BA. Family risk factors versus peer
risk factors for drug abuse. A longitudinal study of an African
American urban community. J Subst Abuse Treat 2000;18:
26775.
4. Christoffersen MN, Soothill K. The long-term consequences of
parental alcohol abuse: a cohort study of children in Denmark.
J Subst Abuse Treat 2003;25:10716.
5. Poulton R, Caspi A, Mine BJ, et al. Association between chil-
drens experience of socioeconomic disadvantage and adult
health. Lancet 2002;360:16405.
6. Ary DV, Tildesley E, Hops H, et al. The inuence of parent,
sibling and peer modeling and attitudes on adolescent use of
alcohol. Int J Addict 1993;28:85380.
7. Wills TA, Vaughan R. Social support and substance use in
early adolescence. J Behav Med 1989;12:32139.
8. Poikolainen K, Tuulio-Henriksson A, Aalto-Setala T, et al.
Predictors of alcohol intake and heavy drinking in early adult-
hood: a 5-year follow-up of 1519-year-old Finnish adoles-
cents. Alcohol Alcohol 2001;36:858.
9. Andersen A. Epidemiological studies on alcohol use in ado-
lescence. (PhD thesis). Copenhagen, Denmark: University of
Copenhagen, 2004.
10. Anda RF, Whiteld CL, Felitti VJ, et al. Adverse childhood
experiences, alcoholic parents, and later risk of alcoholism and
depression. Psychiatr Serv 2002;53:10019.
11. Felitti VJ, Anda RF, Nordenberg D, et al. Relationship of
childhood abuse and household dysfunction to many of the
leading causes of death in adults. Am J Prev Med 1998;14:
24558.
12. Galobardes B, Lynch JW, Davey Smith G. Childhood socio-
economic circumstances and cause-specic mortality in
660 Osler et al.
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

adulthood: systematic review and interpretation. Epidemiol
Rev 2004;26:721.
13. Pensola TH, Martikainen P. Cumulative social class and mor-
tality from various causes of adult men. J Epidemiol Com-
munity Health 2003;57:74551.
14. Claussen B, Davey Smith G, Thelle D. Impact of childhood
and adulthood socioeconomic position on cause specic mor-
tality: the Oslo Mortality Study. J Epidemiol Community
Health 2003;57:405.
15. Nordentoft M, Wandall-Holm N. 10 year follow up study of
mortality among users of hostels for homeless people in Co-
penhagen. BMJ 2003;327:814.
16. Osler M, Andersen AM, Lund R, et al. Revitalising the Met-
ropolit 1953 Danish male birth cohort: background, aims and
design. Paediatr Perinat Epidemiol 2004;18:38594.
17. Green A. The Danish Conscription Registry: a resource for
epidemiological research. Dan Med Bull 1996;43:4647.
18. Munk-Jrgensen P, Mortensen PB. The Danish Psychiatric
Register. Dan Med Bull 1997;44:824.
19. Andersen TF, Madsen M, Jrgensen J, et al. The Danish Na-
tional Patient Register. A valuable source of data for modern
health science. Dan Med Bull 1999;46:2638.
20. Juel K, Helweg-Larsen K. The Danish register of cause of
death. Dan Med Bull 1999;46:3457.
21. Grambsch PM, Therneau TM. Proportional hazard tests and
diagnostics based on weighted residuals. Biometrika 1994;
81:51526.
22. Chenet L, McKee M, Osler M, et al. Alcohol policy in the
Nordic countries. Why competition must have a public health
dimension. BMJ 1997;314:11423.
Social Environment and Substance-related Disorders 661
Am J Epidemiol 2006;163:654661

b
y

g
u
e
s
t

o
n

N
o
v
e
m
b
e
r

2
4
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m