Am J Psychiatry 163:1, January 2006
Toward a Comprehensive Developmental Model
for Major Depression in Men
Kenneth S. Kendler, M.D.
Charles O. Gardner, Ph.D.
Carol A. Prescott, Ph.D.
Objective: The multiple risk factors for
major depression are interrelated through
poorly understood developmental path-
ways. In 2002, the authors presented a de-
velopmental model for major depression
in women. Based on similar methods, they
here present an analogous model for men.
Method: Using data from 2,935 adult
male twins, interviewed twice over a 2–4-
year period, the authors constructed, by
means of structural equation modeling, an
integrated etiologic model for major de-
pression that predicts depressive episodes
over 1 year from 18 risk factors conceptu-
alized as five developmental “tiers” reflect-
ing childhood, early adolescence, late ado-
lescence, adulthood, and the last year.
Results: The best-fitting model, including
six correlations and 76 paths, provided a
good fit to the data, explaining 49% of the
variance in the liability to depressive epi-
sodes. The overall results, similar to those
seen in women, suggest that the develop-
ment of major depression results from the
action and interaction of three broad path-
ways of internalizing symptoms, external-
izing symptoms, and adversity. Childhood
parental loss and low self-esteem were
more potent variables in the model in
men than in women. Genetic risks for ma-
jor depression had a broader spectrum of
action in men than in women. The path-
way to major depression through external-
izing symptoms was not more prominent
in men than in women.
Conclusions: Major depression in men,
as in women, is an etiologically complex
disorder influenced by risk factors from
multiple domains that act in developmen-
tal time. The similarities in etiologic path-
ways to major depression for men and
women outweigh the modest differences.
(Am J Psychiatry 2006; 163:115–124)
Major depression is a paradigmatic multifactorial
disorder, where risk of illness is influenced by a range of
factors including genetic liability, poor parenting, trau-
matic experiences, predisposing personality traits, early-
onset anxiety disorder, poor self-esteem, low social sup-
port, substance misuse, marital difficulties, a prior history
of major depression, and recent stressful life events and
difficulties. The need to organize these diverse risk factors
into an integrated etiologic model to elucidate develop-
mental pathways has been long recognized (1). In 2002, we
published such a preliminary model in female twins (2).
Numerous studies have examined sex differences in the
prevalence and risk factors for major depression (3, 4).
While higher rates in women are consistently reported,
finding robust, replicable differences in risk factors for
major depression in the two sexes has been more difficult.
Furthermore, studies have typically compared only a small
number of risk factors. We are unaware of prior attempts
to compare comprehensive etiologic models for major de-
pression in the two sexes.
This report has two goals. First, we describe a detailed
developmental model for the etiology of major depression
in men that was developed by using methods that parallel
our prior efforts in women (2). Second, we compare the re-
sults of these etiologic models in women and men.
We used data from a two-wave study of male-male and male-
female pairs from the Virginia Twin Registry, formed by a search
of all Virginia birth certificates since 1918. Twins were eligible if
one or both members were successfully located, members of a
multiple birth including at least one male, Caucasian, and born
between 1940 and 1974 (5). Of 9,417 eligible individuals for the
first wave (time 1), interviews were completed, typically by tele-
phone, with 6,814 (72.4%). At least 1 year later, we recontacted the
twins to schedule a second-wave interview (time 2). This inter-
view was successfully completed, mostly face-to-face, with 5,629,
or 82.6% of those eligible. Signed informed or verbal consent was,
respectively, obtained prior to all face-to-face and telephone
interviews. To assess test-retest reliability, 131 members of the
male-male pairs were reinterviewed a mean of 4.4 months (SE=
1.1) after their initial interview.
This report is based on 2,935 members of male-male pairs who
completed both interviews: 1,197 complete pairs and 541 single
twins whose co-twin did not complete both interviews. At the time
2 interview (1994–1998), the subjects had a mean age and years of
education of 37.0 (SD=9.2) and 13.5 (SD=2.7), respectively. The in-
terviewers were clinically trained. Each interview was reviewed
twice for completeness and consistency. The two members of each
twin pair were interviewed by different interviewers.
Our model predicted episodes of major depression in the year
prior to the time 2 interview. Major depression was treated as a di-
Am J Psychiatry 163:1, January 2006
MAJOR DEPRESSION IN MEN
chotomous variable, with the assumption of an underlying nor-
mal liability distribution. In the time 2 interview, twins were asked
about the occurrence in the last year of 15 symptoms reflecting all
DSM-III-R A criteria for major depression. They then aggregated
these symptoms in time, reported the total number of episodes,
and dated, to the month, the onset and offset of each episode. We
examined the first reported episode meeting the criteria unless
there were multiple episodes and the first episode began in the
first 2 months of the year, in which case we took the next reported
episode. The test-retest reliability for last-year major depression
was good: kappa=0.74 (SE=0.08), tetrachoric r=0.96 (SE=0.03).
The variables examined in this study paralleled as closely as
possible those used in our prior investigation of twins from fe-
male-female pairs (2). Exact replication was not possible because
our four-wave study of female-female pairs contained variables
not assessed in our male-male pairs. Also, we had personally in-
terviewed parents of our female-female but not male-male pairs.
As previously (2), we examined 18 predictor variables orga-
nized into “tiers” roughly approximating five developmental peri-
ods: childhood (genetic risk, low parental warmth, childhood sex-
ual abuse, and parental loss), early adolescence (neuroticism, low
self-esteem, early-onset anxiety, and conduct disorder), late ado-
lescence (low educational achievement, lifetime traumas, low so-
cial support, and substance misuse), adulthood (history of di-
vorce and past history of major depression), and the last year
(last-year marital problems, difficulties, and dependent and inde-
pendent stressful life events). (The latter four of these tiers are
conceptual rather than statistical entities as the final 15 variables
in the model are simply consecutively ordered. The first four are
distinct because they are interconnected by correlations—de-
picted by two-headed arrows in the figures—rather than partial
regression coefficients—depicted by one-headed arrows.)
One of these 18 predictor variables, substance misuse, was la-
tent and was constructed, by using a measurement model, from
other observed variables. We here outline briefly each variable;
for further details see our previous article (2).
Genetic risk. Genetic risk was assessed by a composite measure
of the lifetime history of major depression in the co-twin (assessed
at time 1 and time 2) and in the mother and father as assessed by
the family history report of the two twins at time 2 according to the
Family History Research Diagnostic Criteria (6). Parents were sep-
arately divided into three liability categories: history of major de-
pression as reported by neither, one only, or both members of the
twin pair. Co-twins were divided into three categories reflecting
their report of a history of major depression at neither, one, or
both of the personal interviews. To correct for varying base rates
and degree of genetic relatedness in these relatives, we calculated
the modified midrank score for the lifetime history of major de-
pression and adjusted these scores to account for the varying ge-
netic correlation with the proband twin (1.00 for monozygotic co-
twins and 0.50 for dizygotic co-twins and parents). We then took
the mean of the three scores of the co-twin, mother, and father.
Low parental warmth. This variable was assessed by using a
modified version of the Parental Bonding Instrument (7). We took
the mean of up to eight reports from a twin pair with each twin re-
porting on the level of warmth he received and he observed his
twin receiving from both their mother and father. In our study of
female-female twins, this variable was termed “disturbed family
environment” because it included additional measures of family
environment not available in our male-male sample and also in-
cluded reports from the parents of the twins.
Childhood sexual abuse. This was determined from a single
item in the time 1 interview: “Have you ever been sexually abused
or molested?” If a positive response was given, the age at which
this first occurred was recorded. In this report, childhood sexual
abuse was considered present if the age given was prior to 16.
Parental loss. This binary measure was scored as 1 if the twin
reported that one or more parents left the nuclear home through
death, divorce, or parental separation prior to age 17.
Neuroticism. The short-scale (12-item) version from the revised
Eysenck Personality Questionnaire (8) was used to obtain a neu-
roticism score at time 1. Because of its J-shaped distribution, we
scored it as a five-level ordinal measure.
Self-esteem. The full Rosenberg self-esteem scale (9) was ad-
ministered at time 1. It was reverse scored so that higher scores
reflected lower self-esteem.
Early-onset anxiety disorder. This was a binary variable
scored as 1 for subjects with an onset prior to age 18 of panic dis-
order, generalized anxiety disorder, or any form of phobia as as-
sessed at the time 2 interview by using diagnostic criteria outlined
Conduct disorder. We treated conduct disorder as an ordinal
variable that reflected the number of DSM-IV conduct disorder
criteria met prior to age 18 that were endorsed at time 1.
Years of education. Education level was treated as a continu-
ous variable and was assessed at the time 1 interview. It was re-
verse scored to reflect low education.
Lifetime traumas. The number of traumas was reflected by the
number of items reported at the time 1 interview that assessed
exposure to combat, life-threatening accident, natural disaster,
severe injury, physical assault, and being threatened with a
weapon. The distribution was skewed, so it was treated as an or-
Social support. Social support was assessed from the time 1 in-
terview. We took the overall mean of 16 items reflecting the fre-
quency of interpersonal contact, the degree of social integration,
and the quality of the relationships with spouse, twin, children,
parents, other relatives, and friends. This measure, which was
scored to reflect lack of social support, was relatively symmetric
and was treated as a continuous variable.
Substance misuse. This was assessed by using a measurement
model derived from a lifetime diagnosis of DSM-III-R alcohol
abuse or dependence assessed at time 1 or time 2, DSM-IV drug
abuse or dependence assessed at time 2, and nicotine depen-
dence as assessed by a score of ≥7 on the Fagerstrom Tolerance
Questionnaire (10) collected at time 2.
Ever divorced. This binary measure was scored as 1 for men
who reported a lifetime history of divorce or annulment at the
time 2 interview.
Prior history of major depression. This was a binary measure
reflecting the presence or absence of a lifetime history of DSM-
III-R major depression, as reported at either the time 1 or time 2
interview, with an onset at least 8 years before the time 2 inter-
view. This time period was chosen to parallel the time period be-
tween last-year and prior history of major depression used in our
model with female twins (2).
Last-year marital problems. We constructed marital problems
as a three-level ordinal variable using seven items assessing the
level of marital satisfaction in the last year, obtained from the So-
cial Interaction Scale (11), at the time 2 interview.
Last-year difficulties and dependent and independent
stressful life events. We assessed these occurrences, using our
stressful life event measures, in the time 2 interview. Each twin
was systematically asked about the occurrence, at any time in the
preceding 12 months, of 11 “personal” events and four classes of
“network” events, each event being dated to the nearest month
with high interrater reliability (12). The dependence of a stressful
Am J Psychiatry 163:1, January 2006
MAJOR DEPRESSION IN MEN
17. Bentler PM: Comparative fit indexes in structural models. Psy-
chol Bull 1990; 107:238–246
18. Bentler PM, Bonett DG: Significance tests and goodness of fit in
the analysis of covariance structures. Psychol Bull 1980; 88:
19. Steiger JH: Structural model evaluation and modification: an
interval estimation approach. Multivariate Behav Res 1990;
20. Kendler KS, Kuhn J, Prescott CA: The interrelationship of neu-
roticism, sex, and stressful life events in the prediction of epi-
sodes of major depression. Am J Psychiatry 2004; 161:631–
21. Kendler KS, Martin NG, Heath AC, Eaves LJ: Self-report psychiatric
symptoms in twins and their nontwin relatives: are twins differ-
ent? Am J Med Genet Neuropsychiatr Genet 1995; 60:588–591
22. Kendler KS, Pedersen NL, Farahmand BY, Persson P-G: The
treated incidence of psychotic and affective illness in twins
compared to population expectation: a study in the Swedish
Twin and Psychiatric Registries. Psychol Med 1996; 26:1135–
23. Kessler RC, McGonagle KA, Zhao S, Nelson CB, Hughes M, Eshle-
man S, Wittchen H-U, Kendler KS: Lifetime and 12-month prev-
alence of DSM-III-R psychiatric disorders in the United States:
results from the National Comorbidity Survey. Arch Gen Psychi-
atry 1994; 51:8–19