Psychiatric disorders in the relatives of depressed probands I. Comparison of prepubertal, adolescent and early adult onset cases

ELSEVIER Journal of Affective Disorders 42 (1997) 9-22

JOURNAL OF

AFFECTIVE DISORDERS

Research report

Psychiatric disorders in the relatives of depressed probands I. Comparison of prepubertal, adolescent and early adult onset

cases

Richard Harrington a**, Michael Rutter a, Myrna Weissman b, Hazel Fudge a, Christine Groothues a, Diana Bredenkamp a, Andrew Pickles a, Richard Rende b,

Priya Wickramaratne b a Department of Child Psychiatry. Institute ofP,sychiutry, London, UK

b College of Physiciuns and Surpon.s, Cdumhiu University, New York. NY, USA

Received 29 February 1996; accepted 2 I August I996

Abstract

Research with adults suggests that early onset of depression is associated with increased rates of depression among relatives. This paper presents results of a family study that tested the hypothesis that prepubertal depression was associated with a greater familial loading of depression than the postpubertal form, which in turn had a greater familial loading than adult onset depression. Probands were from a child to adult longitudinal study. Psychiatric disorders among relatives were assessed with family interview and family history methods ‘blind’ to all findings regarding the proband. Contrary to expectation, familial rates of depression did not differ significantly between the groups. However, manic disorders tended to be more common among the relatives of postpubertal depressed cases than among the relatives of adult onset cases. Moreover, relatives of prepubertal depressed subjects had higher rates of criminality and family discord than postpubertal subjects. Prepubertal onset depressive disorders appear to be relatively distinct from postpubertal forms. 0 1997 Elsevier Science B.V. All rights reserved

Keywords: Depression; Adolescent; Age; Pubertal; lkelopment

1. Introduction

Depressive disorders with an onset in adulthood tend to run in families (Weissman et al., 1984a). This familial loading has proved to be a valuable

tool in defining the disease phenotype. For instance, family study data suggest that there are at least two distinguishable forms of familial affective disorder, unipolar and bipolar types (Per& 1992). Unfortu- nately, further subdivision of the affective disorder

phenotype has been a difficult task, and as a result many different subtypes of depressive disorder have

* Corresponding author. Department of Child and Adolescent Psychiatry, Royal Manchester Children’s Hospital, Pendlebury,

Manchester. M27 1 HA, UK.

been studied in the hope that they might represent a more homogeneous form of the condition. The ratio-

nale has been that once uniform subtypes are identi-

0165.0327/97/$17.00 Copyright 0 1997 Elsevier Science B.V. All rights reserved.

P/I SOl65-0327(96)00091-2

IO R. Hurringron er ul. / Journal oj’Aj@ctiue Disorders 42 (1997) 9-22

fied then it will be much easier to dissect out specific

genetic and environmental aetiologies. One group that has commanded increasing atten-

tion over the past decade has been the early onset form of the disorder. Interest in this group has been

stimulated by the finding that within samples of

adult probands earlier age of onset of depression

tends to be associated with an increased risk for affective illness among relatives (Weissman et al.,

1987, 1988). Thus, in one of the first papers pointing out this trend, Weissman et al. (1984bl reported that the relatives of probands whose depression had

started before the age of 20 years had higher rates of

major depression than the relatives of probands whose depression had had an onset later in life. Many other

studies of unipolar depressed probands have found that the risk for affective illness among first degree relatives was greater with early than with later age- of-onset cases (Strober, 19921, though the differ-

ences have not always been very large (McGuffin et al., 1988).

Most of these family studies were concerned with an age cutoff that was in the adult range (20 to 51 years). However, family studies of child and adolescent probands with affective disorders (Dwyer and Delong, 1987; Goodyer et al., 1993; Harrington et al., 1993; Kutcher and Marton, 1991; Livingston et al., 1985; Mitchell et al., 1989; Puig-Antich et al., 1989; Strober et al., 1988; Todd et al., 1993) have shown that there are often high rates of affective disturbance among first degree relatives (FDRs). In- deed, several authors have suggested that these disorders might represent the most severe form of the condition, with a particularly high familial loading (Puig-Antich et al., 1989; Strober et al., 1988: Todd et al., 1993).

If it is the case that the familial loading for affective disorders is especially high in very early onset cases then there would be a number of implica-

tions. One of the most important is the idea that this increased familial loading could reflect a higher genetic risk, and that therefore these families would be more appropriate for genetic studies than the families of probands with an onset in adulthood (Todd et al., 1993). However, genetic investigations such as twin studies will be expensive and difficult to undertake

in this age group, if only because of the relative rarity of severe depressive disorders in prepubertal

children (Harrington, 1993). Moreover, interpretation

of the results will not always be straightforward. For instance, it may be difficult to establish twin regis-

ters that generate representative samples of twins for child psychiatric studies. In addition, although the

equal environments assumption seems to be approxi-

mately correct for normal traits and many adult

psychiatric conditions (Kendler, 19931, it has not yet

been adequately demonstrated for child psychiatric

disorders (Goodman, 1991; Rutter et al., 1990; Rut-

ter and Redshaw, 1991; Simonoff et al., 1994). So, it will be important that maximum use is made of

family study data in order to generate specific hypotheses that can be tested in twin or adoption

designs. Indeed, modem methods of biometrical

model fitting, which are now widely used in such designs, require that competing hypotheses are clearly articulated before the data are analyzed (Kendler,

1993).

Two important issues have yet to be tackled in family studies of early onset depressive disorders. First, it is still not known whether familial loading for affective conditions is actually any higher in prepubertal onset cases than it is in adolescent or

adult forms of the disorder. Thus, so far as we know there have been no published studies in which depressed probands with an onset in childhood, adolescence or adulthood have been simultaneously compared in respect of familial loading for affective disorders.

The second issue concerns the extent to which the family study data are consistent with possible genetic or environmental modes of transmission. Most fam-

ily studies of child or adolescent probands with depression have concentrated on establishing familial rates of psychiatric disturbance, and very few have specifically used these data to test hypotheses about genetic or environmental influences.

This paper and its companion present results of two complementary family studies that address these

issues. Both studies used similar methodologies. The British study compares rates of affective disorders among the relatives of depressed probands known to have had prepubertal, adolescent or early adult on- sets of depressive disorders. The American study

extends the findings of the British study by examining the impact of age of onset from early adolescence through to middle age.

R. Hurrington et ai./Journol oj’A$xtive Dim-ders 42 (1997) 9-22 II

In the present paper the results of the British

study are described. The main aim was to establish whether or not there were systematic developmental trends in the familial loading for psychiatric disor-

ders. In particular, we aimed to test the hypothesis

that depressive disorders that had a very early onset

(prepubertal) had a significantly higher familial load-

ing of affective disorders than disorders with an

onset in adolescence (pubescent or postpubertal), or

early adult life. The study also aimed to establish whether or not there were age trends in other factors

that might be relevant to aetiology, such as family

discord.

2. Methods

2.1. Selection of depressed probands and families

The British study was based on the Maudsley follow-up into adulthood of 80 depressed child

probands and 80 non-depressed child psychiatric controls (Harrington et al., 1990; Harrington et al., 1991). At follow-up, on average 18 years after the initial contact, outcome information was obtained on 62 depressed subjects and 69 controls ‘blind’ to the childhood data (Harrington et al., 1993; Harrington et al., 1990). We interviewed at least one first degree relative in 111 (85%) of these families and it is this group who formed the basis of the family study.

The longitudinal design of our study allowed us to use the proband’s affective status in both childhood and adulthood (‘adulthood’ was defined as age 17

years or more - see Harrington et al. (1990)) to generate four proband groups: (1) prepubenal onset depressive disorder (onset age 6 to 14 years) (n = 18); (2) pubescent or postpubertal depressive disor-

der, hereinafter referred to as ‘postpubertal’ (onset age 11 to 16 years) (n = 33); (3) adult onset depressive disorder (i.e., child psychiatric controls who had an episode of RDC depressive disorder in ‘adulthood’) (n = 16); and (4) a child psychiatric control group who had not been depressed in either childhood or adulthood, but who had attended for non-depressive disorders such as conduct disorder and anxiety states (n = 44). Pubertal status was assessed routinely at the time of the index episode by history

or physical examination. So far as we could tell from

the charts, the adolescent cases had all had an onset of depression at or after puberty.

Pubertal status was chosen as a marker of development, rather than age, because (a) epidemiological

studies have shown that the increase in rates of

depression during adolescence is more closely re-

lated to pubertal status than age (Rutter et al., 19701, (b) in our follow-up study, there were significant

differences between the outcomes of pre- and post-

pubertal depressed cases (Harrington et al., 1990).

2.2. Diagnosis of depression in childhood

The follow-up study was based on the clinical

data records (‘item sheets’; see Thorley, 1982, 1986, 1987) of child and adolescent patients who attended

the Maudsley Hospital during the late 1960s and early 1970s. As described previously (Harrington et al., 1990) the item sheets were used to identify 80 child patients with an operationally defined depressive syndrome and 80 non-depressed child psychiatric controls. The reliability of the psychiatrists who compIeted the item sheets was unknown. However, Goodman and Simonoff ( 199 1) reported that Mauds- ley psychiatrists’ ratings of the item sheets can be of acceptable reliability for research purposes. More- over, there were several reasons for thinking that the depressive syndrome had validity. First, independent re-diagnosis of the charts (Harrington et al., 1990) showed that most of the depressed group met criteria for a Research Diagnostic Criteria (RDC) depressive diagnosis (Spitzer et al., 1978). Second, subjects with the depressive syndrome resembled cases iden- tified as depressed in other studies in a number of key respects, such as sex ratio and degree of overlap with conduct problems (Harrington et al., 1990, 1991). Third, two independent studies have shown that the depressive syndrome has predictive validity, to the extent that it predicts depressive disorder in adulthood (Harrington et al., 1990; Zeitlin, 19861. Finally, the validity of the syndrome as a tool for examining familial patterns of disorder is shown by our previous finding that relatives of children with the syndrome have significantly higher rates of depressive disorder than non-depressed child psychiatric controls (Harrington et al., 19931.

12 R. Hurrington et al. / Journal of’Aj$xriur Disorders 42 (1997) 9-22

2.3. Assessment of family discord in childhood

Ratings of intrafamilial discord were derived from

the Maudsley charts, which contained descriptions of

family life and relationships at the time of the proband’s referral. A section on family life and

relationships was a standard part of the Maudsley

history at this time. These descriptions were exam-

ined for evidence of (1) discord between the parents;

(2) harsh parenting and/or criticism directed to-

wards the proband by the predominant parent, usually the mother. Ratings were based on the criteria

used in the Camberwell Family Interview, which

have been used in previous studies of the association between psychiatric problems in children and their

parents (Quinton and Rutter, 1985; Rutter and Quin-

ton, 1984). Assessments were made by a rater who was blind to the family study findings.

2.4. Diagnosis of psychopathology in adult probands

and relatives

The methods used to make psychiatric diagnoses on the probands in adulthood and on adult relatives have been described in detail elsewhere (Harrington et al., 1988, 1990, 1993). Briefly, psychiatric status of interviewed subjects was assessed using a modi- fied version of the Schedule for Affective Disorders and Schizophrenia-Lifetime version (SADS-L) (Spitzer and Endicott, 197.5), and diagnoses were made according to the RDC (Spitzer et al., 1978). Since we have previously shown that the reliability of lifetime diagnoses is higher when there is impair-

ment or help seeking (Harrington et al., 19881, only depressive disorders (major depressive disorder (MDD) or minor RDC depression) that had led to one or other of these states were included.

If it was not possible to interview the subject directly, then information about this relative was obtained from other sources. We distinguished between two types of non-interview data about relatives. (a) ‘High quality’ non-interview data (‘informant’ data) were obtained either from a detailed interview with one informant who knew the subject

well for most of his/her life, and/or from a ques-

tionnaire completed by those subjects unwilling to have a face to face interview. (b) ‘Pedigree’ data were less detailed data about the mental health of

relatives obtained during the construction of a family

pedigree using the Family History Method (Gershon and Guroff, 1984). All data on second degree rela-

tives (SDRS) were collected using the Family His-

tory Method - Research Diagnostic Criteria @I-I-

RDC). Criminal records on FDRs were obtained from the

Central Criminal Records Office in London. It was

not possible to obtain records on SDRs because of

insufficient identifying details. Accordingly, data on

criminality among SDRs were collected during the

FH-RDC.

Diagnoses in adult probands and their relatives

were made ‘blind’ to the childhood data. To ensure comparability between the British and

American interview techniques, a research adminis-

trator from the American group visited London to train the British interviewers in the use of the SADS-

L. A pilot study was then conducted in which the two techniques were compared systematically in a test-retest design. There was good agreement (kappa

more than 0.7) between the two techniques on most lifetime psychiatric diagnoses (Harrington et al., 1988).

2.5. Statistical methods

Tabulations, descriptive statistics and t-tests of differences in sample means were obtained using SPSS-PC (Norusis, 1990). Significance levels quoted for tabulated data were based on exact tests using EGRET (SERC, 1990). For analysis of risk to relatives, Cox proportional hazards models (Cox, 1972) were also used to provide estimates of the relative risk of disorder adjusted for both the observed time that each relative was exposed to the risk and other confounding variables. These were estimated using BMDP-PC (Software, 1990).

3. Results

3.1. Characteristics of probands and relatives

Table 1 shows that the two childhood depressed groups did not differ significantly in respect of de- mographic variables (other than, of course, age), comorbidity with behavioural disturbance, or indices

R. Hawingtan er al./Journul ofAj@ctiue Disorders 42 (1997) 9-22 13

Table I Characteristics of probands

Proband group Significance

Pre-pubertal Post-pubertal Adult Never 1 2 3

depressed depressed depressed depressed

No. of probands 18 33 16 44

Age first depression mean 10.7 14.2 20.1

(4.3) I

< 0.001 < 0.001

SD (2.3) (1.3)

Age at follow-up mean 29.2 32.5 29.7 30.0 NS < 0.01 <O.OOI

SD (3.5) (2.1) (2.2) (3.4)

Female, n (o/o) 6 (33) 18 (54) 9 (56) 1709) NS NS NS

Psychiatric contact prior to index 4 (22) 14 (42) 5 (33) 14 (34) NS NS NS

attendance a, n (%I

Conduct disorder in childhood, n (8) 3 (17) 9 (27) 4 (25) IO (23) NS NS NS

Manual social class b, n (o/n) 16 (89) 21 (64) 13 (81) 34(77) NS NS NS

No. adult psychiatric mean 0.8 2.1 1.3 0.02 <O.oOl NS NS

hospitalizations

SD (2.5) (3.8) (2.5) (0.15)

Adult antisocial personality, n (%I 6 (33) 6 (18) 3 (19) lO(23) NS NS NS

Adult general social dysfunction ‘, n (%) 4 (22) 14 (45) 9 (56) II (26) NS NS NS

Significance: I = all depressed probands (i.e., pre-, post and adult) versus never depressed.

2 = prepubertal depressed versus postpubertal depressed.

3 = Adult depressed versus postpubertal depressed.

a Missing dam on 4 cases; b British Registrar General classification: labouring and blue collar jobs; ’ Missing data on 4 cases, ratings based

on the Adult Personality Functioning Assessment (Hill et al., 1989).

of severity of psychopathology in adulthood such as social dysfunction (see Hill et al., 1989) or number

of psychiatric hospitalizations. As Table 2 indicates, the relatives of the four

proband groups were in general well matched on age, sex, and source of information. However, compared with the other groups, information on FDRs of probands who had never been depressed was more

likely to have come from the FH-RDC method.

3.2. Lifetime rates of disorders among relatives

Table 3 shows the lifetime rates of various RDC disorders, and of criminality, in relatives. Compared with non-depressed controls, depressed cases had significantly (P < 0.05) increased rates of major depressive disorder among both first and second degree relatives. However, contrary to our hypothesis, familial loading of depression did not increase systematically with decreasing age. Indeed, as the table shows, the relatives of probands whose depression had started prepubertally tended to have lower rates

of depressive disorder than the relatives of probands with an adolescent onset. By contrast, criminality was significantly more common among the FDRs of prepubertal depressed cases as compared to postpubertal depressed cases.

These results could be confounded by any one of a number of factors. For example, whether for sub- stantive or methodological issues, there is evidence that individuals born in recent years report higher rates of depression than those born earlier in the century (Fombonne, 1995; Klerman, 1988). To ad- just for the effects of these potential confounding variables, these analyses were repeated using multi- variate proportional hazards models with the following covariates added to the equation: sex of relative, year of birth, quality of information (direct interview, high quality informant or pedigree data) severity of disorder in the proband (a compound variable, comprising admission to hospital as a child. previous psychiatric care in childhood, and number of hospital admissions in adult life), comorbidity with conduct disorder, sex of proband, degree of relative, and

14 R. Hurrington et ul./ Journul ofAtiective Disorders 42 (1997) 9-22

Table 2

Characteristics of relatives

Proband group Significance

Pre-pubertal Post-pubertal Adult Never I 2 3


First degree relatives No. 67 I13 60 159 Age mean 45.3 49.6 45.6 44.8 NS NS NS

SD (15.5) (15.9) (15.5) (44.8) Female, n (%) 34 (51) 59 (52) 27 (45) 79 (50) NS NS NS

Information. n (6)

Interview

Informant

FH-RDC

48 (72) 80 (71) 46 (77) 105 (66)

II (16) 26 (23) IO (17) 24 (15)

8 (12) 7 (6) 4 (7) 30 (19) < 0.01 NS NS

Second degree relatives No. 109 175 107 244 Age mean 64.6 64.5 63.0 64. I NS NS NS

SD (14.5) (17.4) (17.2) (16.4)

Female, n (%o) 54 (49) 89 (51) 58 (54) 131 (54) NS NS NS

Significance: I = all depressed probands (i.e., pm, post and adult) versus never depressed.


3 = adult depressed versus postpubertal depressed.

whether of proband’s generation. The results were not materially different from those obtained using the simpler proportional hazards techniques.

We have previously noted the weaknesses of our childhood diagnoses of depression and accordingly rediagnosed the cases by applying the Research Di-

Table 3

Lifetime rates (%1/o) of disorders in first and second degree relatives

Proband group Relative risk a

Pre-pubertal Post-pubertal Adult Never 1 2 3


Major depression FDR 14.9 16.8 13.3 7.5 2.2 c I .o 0.8

SDR 4.6 7.4 7.5 0.4 9.0 c 0.5 I.0

Anxiety disorder b FDR 10.4 10.6 15.0 11.3 0.8 1.1 1.6

SDR 0.9 1.7 0.9 0.8 2.1 0.5 0.5

Alcohol disorder FDR 13.4 7. I 6.7 9.4 0.7 2.1 I.0

SDR 8.3 4.6 6.5 3.7 1.2 1.8 1.5

Criminal record FDR 23.9 9.7 18.3 11.3 0.8 2.7 ’ 2.0

SDR 2.8 2.3 2.8 2.9 0.8 1.2 1.2

a Proportional hazards model. Numbers of first degree relatives (FDRs) and second degree relatives (SDRS) as in Table 2.

b Generalized anxiety phobia or panic disorder. c P < 0.05.

I = all depressed probands (i.e., pre-, post and adult) versus never depressed.



R. Hurrington et d/Journal ofAj$xtivr Disorders 42 (1997) 9-22 15

agnostic Criteria to the charts (Harrington et al., 1990). All the analyses described above were therefore repeated after the proband’s status was rede- fined according to the RDC chart diagnoses. The results were very similar to those obtained using the original proband designations.

4. Tests of hypotheses about possible mechanisms

It seemed then that early onset depressive disorders were associated with a family history of depression. However, contrary to what had been expected, prepubertal cases tended to have lower familial rates of depression than postpubertal cases and significantly higher rates of criminality among FDRs. Two questions arose from these findings. First, what factors could account for these links between psychopathology in families and depression in off- spring? Second, to what extent did these factors differ between pre- and postpubertal onset cases?

Family study designs do not provide a clear testing of possible genetic and environmental mechanisms. Nevertheless, as we noted above it is important to use family study data to test expectations and hypotheses that derive from genetic and environmental theories. We used three main strategies to test these hypotheses.

4.1. Type of affective disorder in relatives

The first set of expectations concerns the types of affective disorders in relatives. Research on adults with affective disorders suggests that genetic factors are strongest for bipolar forms (McGuffin and Katz, 1986) and for recurrent unipolar disorders (Kendler et al., 1993b), and less strong for mild unipolar conditions (McGuffin and Katz, 1986). It follows that early onset depressions with the highest genetic liability should have the greatest familial loading of those forms of affective disorder that are thought to be most genetic, namely bipolar disorder and recurrent major depression.

Table 4 shows the most severe types of affective disorders that occurred in the relatives of the four proband groups. Overall, there were few differences between the groups. However, as the table shows, the rate of mania or hypomania in the families of postpubertal depressed cases tended to be higher than in the families of both prepubertal cases ( ,y2 = 2.8, df = 1, P = 0.09) and adult depressed subjects ( x2 = 4.7, df = 1, P = 0.03). By contrast, familial rates of minor depressive disorders were very similar across all four proband groups.

A related issue concerns the extent to which depressive disorders in relatives are associated with non-depressive disorders. It cannot be said that the

Table 4

Most severe lifetime affective disorder in all relatives

Proband group Significance ’

Pm-pubertal Post-pubertal Adult Never I 2 3


Never depressed % 81.3 77.4 83.8 89. I

n (143) (223) (140) (359)

Minor depression only % 10.2 IO.1 6.6 7.4

n (18) (29) (II) (30)

Major depression, < 0.001 NS NS

not recurrent % 2.8 2. I 3.6 I.5

#I (5) (6) (6) (6)

Recurrent major depression % 5.1 7.6 6.0 I .5

n (9) (22) (IO) (6) Mania or hypomania % 0.6 2.8 0 0.5

n (I) (8) (0) (2)

a xz tests, df = 4.

I = all depressed probands (i.e., pre-, post and adult) versus never depressed.



16 R. Hurrington et ul./JournuI c$AJjective Disorders 42 (1997) 9-22

Table 5

Lifetime rates of non-depressive disorders among first degree relatives (FDRs) with major depression

Proband group Significance a

Pre-pubertal Post-pubertal Adult Never I 2 3


(N. FDRs with major depression) (IO) (19) (8) (II)

Anxiety disorder (%) 10.0 26.3 62.5 18.2 NS NS NS

Alcohol disorder (%o) 50.0 5.3 25.0 9.1 NS < 0.01 NS

Criminal record (W) 50.0 0.0 12.5 9.1 NS <O.OOl NS

a x’ tests, df = I.

1 = all depressed probands (i.e., pre-, post and adult) versus never depressed.



effects of genes and/or environmental influences on the association between adult depression and comor- bid non-depressive disorders are well understood. However, there is evidence from twin studies that unipolar depressed adults with alcoholic or socio- pathic relatives are distinct from those without (Numberger and Gershon, 19841, whereas major depression and generalized anxiety disorder may share genetic risk factors (Kendler et al., 1992). The question therefore arose as to whether the depressed relatives of pre- and postpubertal depressed children differed in the extent to which they suffered comor- bid disorders such as alcoholism, criminality and anxiety.

Table 5 shows that they did. The depressed FDRs of prepubertal depressed children had significantly greater lifetime rates of alcohol use disorder and criminality than the depressed relatives of postpubertal depressed children. By contrast, the risk of an

anxiety disorder among the depressed FDRs of depressed probands increased steadily with age of onset of depression in the proband, being highest in the relatives of probands with an onset in adulthood (x2=6.1, df= 2, P < 0.05). The number of SDRs with MDD was too small to permit meaningful analysis.

4.2. Familial aggregation of possible environmental variables

The next expectation concerns the pattern of associations with variables that are thought to indicate family environmental factors. It will be appreciated that genetic and environmental factors will fre- quently be correlated. For instance, recent research suggests that many variables that have traditionally been viewed as environmental risk factors, can have a genetic predisposition (Kendler et al., 1993a;

Table 6

Probands’ family environment at the time of the index episode

Proband group

Pre-pubertal Post-pubertal

depressed depressed

(N. probands) (18) (33) Parental discord (%) 72 45

Maternal criticism and/or hostility (%) 78 36

Significance a

Adult Never 1 2 3

depressed depressed

(16) (44) 62 64 NS 0.07 NS

37 59 NS < 0.01 NS

a x’ tests, df = I.

I = all depressed probands (i.e.. pm-, post and adult) versus never depressed.



R. Hurringmn Ed nl./Journol oj’A&ctiue Disorders 42 (1997) 9-22 17

Plomin and Bergaman, 1991). The presence of sup- posed environmental factors in the families of depressed young people does not therefore prove that they caused the disorder. However, it is clearly the case that the absence of these environmental risk factors weakens the hypothesis of an environmental aetiology.

There are many possible environmental risk factors that have been linked to depression in young people (Harrington, 1993). We studied intrafamilial discord because (1) the charts contained relatively good data on intrafamilial relationships; (2) numer- ous studies have shown that family discord is an important risk factor for many types of child psychiatric disorder (Wolkind and Rutter, 1985), as well as depression (Asamow et al,, 1993; Hammen, 1991; Schwartz et al., 1990); and (3) there is some evidence from non-genetic designs that family discord represents a significant environmental influence on psychopathology in children (Rutter, 1985).

set cases are mainly due to genetic influences then there would be the following expectations; (1) since adverse family environments may not persist into adult life, continuities to depression in adulthood should be weaker in the prepubertal than in the postpubertal group; (2) that in the postpubertal depressed group, continuity to depression in adulthood will be driven at least in part by a family history of depression.

Table 6 shows the association between the two

We reported in an earlier publication from this study that continuity to major depression in adulthood was significantly lower in prepubertal probands than in postpubertal cases (Harrington et al., 1990). However, contrary to our second expectation, within the postpubertal depressed proband group the rate of MDD in the relatives of probands who did develop MDD in adulthood (19/132 or 14.4%) was not significantly different from that in the relatives of probands who did not go on to develop major depression in adulthood (13/156 or 8.3%) ( x2 = 2.7, P = 0.1).

measures of family discord, maternal criticism/hostility directed towards the child and discord between parents, and depression in the proband. Maternal hostility was significantly more common in the families of prepubertal depressed probands than in the families of postpubertal cases. This association was not the result of overlap with conduct disorder, which tended to be less common among prepubertal than postpubertal depressed cases (17% vs. 27%; P = 0.4). Nor was it a result of an association between depression in families and maternal discord. Thus, for the whole sample of 1034 relatives, the rate of MDD in relatives in families where there was maternal criticism of the proband during the proband’s childhood was 7.4%, compared with 6.9% in relatives in families without maternal criticism. However, criminality was significantly more common in families where there was maternal criticism (9.4%) than in families where there was not (4.5%) ( x2 = 9.6, P < 0.01).

5. Discussion

5.1. Methodological limitations

4.3. Continuities of depressive disorder over time

The present study combined a child to adult follow-up with a family genetic design. This combina- tion provided a unique opportunity to examine the linkages between age of onset of depression and familial loading of the disorder. In particular, it was possible to study the power of family history to predict continuities to adulthood. Moreover, the adult outcome data on the child probands enabled us to identify a control group who were not depressed in either childhood or adult life. This was important because of the suggestion from previous family studies of depressed children that apparently non-depressed child psychiatric control groups could be- come contaminated with cases who went on to develop depression later in life (Puig-Antich et al., 1989).

The final set of hypotheses stem from expecta- Nevertheless, in interpreting the results the fol- tions about continuities over time. If it is argued that lowing methodological limitations should be borne prepubertal onset cases of depression are mainly due in mind. First, the study was based on criteria for to adverse family environments and postpubertal on- childhood depression that were devised many years

18 R. Hurringron et d/Journal oj’A&ctiue Disorders 42 (1997) 9-22

ago (Pearce, 1978) and which differ in several respects from modem diagnostic concepts (Harrington et al., 1990). In particular, fewer symptoms are required and some of these symptoms, such as school

refusal, are only found in children. It cannot there-

fore be guaranteed that the results will apply to

research based on criteria such as DSM-IV, although there seems to be much overlap between the two

constructs (Harrington et al., 1990).

Second, the reliance of this study on charts and clinical data sheets for information about the

proband’s childhood means that some of our key measures were very crude. For example, there is the

problem of deciding whether the absence of evidence of discord in the charts meant that it was truly absent or whether it was simply the case that the clinician

had failed to question the family properly. Similarly, it is possible that some of our postpubertal depressed

cases had had an onset of depression prepubertally (although the information in the charts suggested that all of them were indeed postpubertal at the time of onset). We should emphasize, however, that all these factors will tend to reduce the chances of finding significant differences between groups.

Third, the family-history method will undoubtedly underestimate the rate of depression in non-interviewed relatives. In addition, there is the danger that the psychiatric status of the informant. usually a relative, will bias the results. Fourth, the results of our clinic-based study may not generalize to cases seen in the community. Indeed, it has been known for a long time that family factors affect referral to child psychiatrists (Shepherd et al., 1971) and as is the case in all clinical studies there is no way of telling how referral biases could affect the results. Fifth, our study did not have a normal control group, and thus will probably have underestimated the importance of non-specific etiological factors. Finally, we note that with small numbers of probands the power of this study to detect differences between groups was limited. For instance, the minimum relative risk detectable at the five per cent level for a disorder in the relatives of prepubertal cases, as compared to postpubertal cases, was more than 2.0 (see Table 3). Lack of power may have been com- pounded by the fact that proportional hazards models do not take into account the non-independence of relatives in family study analyses.

5.2. Early versus adult onset depressions

The relationship between age of onset of depres-

sive disorder and a family history of depression may

be more complex than has been assumed up to now. We noted in the introduction to this paper that many

studies of adults with depressive disorders have found

that early adult onset of depression, usually defined as under 40 years or sometimes less than 20 years

(Strober, 1992; Weissman et al., 1984b), was associ-

ated with a greater familial loading of depression than an onset in middle or old age. Some of our

results are consistent with this literature, especially the finding that the risk of manic disorders tended to be higher in the families of postpubertal onset de-

pressed cases than in the families of cases of depres-

sion with an onset in adulthood. The implication is that adolescent onset depressive disorders can repre-

sent an especially severe form of the condition. This suggestion is compatible with the finding that the prognosis of these disorders is poor, with a significant risk in clinical samples of the development of

mania (Strober et al., 1993). However, our findings also suggest that this in-

creased risk does not extend in a simple linear fashion down to prepubertal onset cases. Such cases did not have a greater familial loading of affective disorders than postpubertal forms. If anything, the risk of affective disorders was a little lower than in postpubertal cases. Rather, prepubertal onset of depression was associated with a family history of criminality and alcoholism, and with greater comorbidity among depressed relatives. These findings are strikingly similar to those reported by Puig-Antich et al. (Puig-Antich et al., 1989) who found that prepubertal onset of major depression was associated with ‘familial comorbidity’ of depression and alcoholism. They are also consistent with the findings of Mitchell et al. (1989) who reported that drug abuse histories were significantly more common in the mothers of preadolescents with major depression than in the mothers of depressed adolescents.

By contrast, postpubertal onset of depressive disorder was associated with a much more specific familial loading of affective disorders, to the extent that no disorder other than depression was increased among first degree relatives. It should be noted, however, that pre- and postpubertal-onset depres-

R. Hurringron er al./Jotm~ul r)j’Affective Disorder.~ 42 (199719-22 19

sions resembled each other in one important respect:

in comparison with the non-depressed child psychiatric controls, both groups showed increased familial rates of depression.

strongly associated with our proxy environmental variable, family discord.

At first sight, our results would seem discrepant with the report that parents with early onset major

depression were more likely to have children with

early onset ( < 13 years) major depression (Weiss-

man et al., 19881. However, it must be borne in mind

that our ‘bottom up’ study was based on child

probands, whereas the high risk study of Weissman

et al. (Weissman et al., 1988) was based on the

children of adult probands. The mechanisms leading

to childhood depression may not be the same in the

two groups.

7. Implications for genetic studies

Of course, none of these findings in themselves proves, or indeed disproves, genetic or environmen-

tal transmission. However, on the basis of these

results it is possible to articulate a number of hy-

potheses that can be tested in more definitive investi-

gations such as twin studies.

6. Mechanisms

What are the reasons for the familial dissimilari- ties between pre- and postpubertal onset depressive disorders that were found in the present study?

Clearly, in view of the relatively crude nature of our childhood assessments, it is first important to con- sider whether the differences between the two groups

could be artifactual. In particular, could they be induced by differences in the reliability of the base- line assessments, such that the prepubertal group

contained fewer ‘true’ cases of depression? Such an artifact could explain the luck of an association with a given correlate, but it is an unlikely explanation for the positive associations. Indeed, the differences between pre- and postpubertal onset cases of depressive disorder were defined not so much by a reduced familial loading for depression as by increased familial rates of other problems, notably criminality and family discord. Moreover, the prepubertal group did have a significantly increased familial loading of depression when compared with the non-depressed

control group.

First, our family study findings would be compatible with the idea that postpubertal-onset depressive

disorders might have a higher heritability than prepu-

bertal-onset disorders. This hypothesis may seem counterintuitive, but there are paradigms in the medi- cal literature of heritability increasing with age. For instance, studies of identical twins have shown that

concordance is considerably higher in the maturity- onset non-insulin dependent form of diabetes than in the juvenile-onset insulin-dependent variety (Leslie and Pyke, 1980). Moreover, Thapar and McGuffin (1994) found that depressive symptoms rated on a questionnaire showed higher heritability in adolescence than in childhood. The literature on psychopathology is sparse, but there is some evidence

that the role of shared environment in conduct disorder decreases with age (Simonoff et al., 1994).

However, an alternative hypothesis also requires testing in genetic designs, namely that the genetic factors that influence risk for depressive disorder in postpubertal cases are only weakly related to those that influence the risk for depression in prepubertal cases. It could be, for instance, that these factors are much more closely linked to alcoholism and/or criminality. In other words, prepubertal onset depressive disorder may be just as heritable as postpubertal or adult forms, it is just that the genotype is differ-

ent.

Rather, the data suggest that there may be real The second set of implications concern the role of

developmental differences between pre- and postpu- family environments. In classic twin models, genetic

bertal onset cases in the mechanisms leading to and environmental risk factors are treated as ‘latent’

depression. Compared with prepubertal cases, post- variables whose presence is inferred from correla-

pubertal forms of depressive disorder had a more tions between relatives. The inclusion of specified

specific association with a family history of depres- risk factors permits the development of more precise

sion and stronger temporal continuities to major models and greatly increases statistical power (Ken-

depression in adulthood. Prepubertal cases were more dler. 1993). Our findings suggest that family discord

20 R. Hurrington et ul./Journal oj’Ajfective Disorders 42 (1997) 9-22

should be specified as a family environmental risk factor in genetic studies of early onset depressive disorders, especially very early onset forms. How- ever, it should also be noted that since there was an association between parental discord and a family history of criminality, the assumption that family discord is purely an environmental risk factor may be incorrect. It could be for instance that one or more temperamental variables predispose parents to criminality and discordant relationships, but in children lead to depression. And, of course, genetic studies will also need to measure non-familial environmental factors.

The final implication concerns the overall design of genetic investigations of early onset depressive disorders. There are now a number of different twin designs available to study the genetics of child psychiatric disorders (Rutter et al., 19901, many of which incorporate multiple occasions of measure- ment. Our data underline the importance of using these designs to cover the transition period of early adolescence, when there may be important changes in genetic and environmental risk factors. It will be vital to follow the same group of individuals over the transition, as well as examining heritability across different ages.

8. Conclusions

The relationship between familial loading and age of onset may be more complex than previously thought to be the case. Future genetic investigations will need to take into account the possibility that prepubertal onset depressive disorders are relatively distinct from postpubertal onset cases. Developmen- tal stage at the time of onset may be a useful marker of heterogeneity.

Acknowledgements

This study was supported by the MacArthur Foundation Network on Risk and Protective Factors in the Major Mental Disorders.

References

Asamow, J.R., Goldstein, M.J., Tompson, M. and Guthrie, D.

(1993) One-year outcomes of depressive disorders in child

psychiatric inpatients: evaluation of the prognostic power of a

brief measure of expressed emotion. J. Child Psychol. Psychia-

try 34, 129-137.

Cox, D.R. (1972) Regression models and life tables (with discus-

sion). J. R. Stat. Sot. 34, 187-220.

Dwyer, J.T. and Delong, R.G. (1987) A family history study of

twenty probands with childhood manic-depressive illness. J.

Am. Acad. Child Adolesc. Psychiatry 26, 176- 180.

Fombotme, E. (199.5) Depressive disorders: time trends and puta-

tive explanatory mechanisms. In M. Rutter and D. Smith

(I%.), Psychosocial Disorders in Young People: Time Trends

and Their Origins. Wiley, Chichester, pp. 544-615.

Gershon, ES. and Guroff, J.J. (1984) Information from relatives:

diagnosis of affective disorders. Arch. Gen. Psychiatry 41,

173-180.

Goodman, R. (1991) Growing together and growing apart: the

non-genetic forces on children in the same family. In P.

McGuffin and R. Murray &is.), The New Genetics of Mental

Illness. Butterworth-Heinema, Oxford, pp. 212-224.

Goodman, R. and Simonoff, E. (1991) Reliability of clinical

ratings by trainee child psychiatrists: a research note. J. Child

Psychol. Psychiatry 32, 55 l-555.

Goodyer, I.M., Cooper, P.J., Vize, C. and Ashby, L. (1993)

Depression in 11 to 16 year old girls: the role of past parental

psychopathology and exposure to recent life events. J. Child

Psychol. Psychiatry 34, 1103-I 115.

Hammen, C. (1991) Depression Runs in Families. The Social

Context of Risk and Resilience in Children of Depressed

Mothers. Springer, New York, NY.

Harrington, R.C. (1993) Depressive Disorder in Childhood and

Adolescence. Wiley Series on Studies in Child Psychiatry.

Wiley, Chichester.

Hanington, R.C., Hill, J., Rutter, M., John, K., Fudge, H., Zoccol-

illo, M. and Weissman, M.M. (1988) The assessment of

lifetime psychopathology: A comparison of two interviewing

styles. Psychol. Med. 18, 487-493.

Harrington, R.C., Fudge, H., Rutter, M., Pickles, A. and Hill, J. (1990) Adult outcomes of childhood and adolescent deptes-

sion: 1. Psychiatric status. Arch. Gen. Psychiatry 47, 465-473.

Harrington, R.C., Fudge, H., Rutter, M., Pickles, A. and Hill, J.

(1991) Adult outcomes of childhood and adolescent depression: II. Risk for antisocial disorders. J. Am. Acad. Child

Psychiatry 30. 434-439.

Hanington, R.C., Fudge, H.. Rutter, M., Bredenkamp, D.,

Groothues, C. and Pridham, J. (1993) Child and adult depression: a test of continuities with data from a family study. Br. J.

Psychiatry 162, 627-633.

Hill, J., Harrington, R.C., Fudge, H., Rutter, M. and Pickles, A.

(I 989) Adult Personality Functioning Assessment (APFA): an

investigator based standardized interview. Br. J. Psychiatry 155, 24-35.

Kendler, KS. (1993) Twin studies of psychiatric illness. Current

R. Hurrington et ul./Journul oj’Aff>ctive Disorders 42 (1997) 9-22 21

status and future directions. Arch. Gen. Psychiatry 50, 905-

915.

Kendler, KS., Neale, MC., Kessler, R.C., Heath, A.C. and Eaves,

L.J. (1992) Major depression and generalized anxiety disorder.

Same genes, (partly) different environments? Arch. Gen. Psy-

chiatry. 49, 716-722.

Kendler, K.S., Neale, M., Kessler, R., Heath, A. and Eaves, L.

(1993a) A twin study of recent life events and difficulties.

Arch. Gen. Psychiatry 50, 789-796.

Kendler, K.S., Neale, M.C., Kessler, R.C., Heath, A.C. and Eaves,

L.J. (1993bI A longitudinal twin study of l-year prevalence of

major depression in women. Arch. Gen. Psychiatry 50, 843-

852.

Klerman. G.L. (1988) The current age of youthful melancholia.

Evidence for increase in depression among adolescents and

young adults. Br. J. Psychiatry 152, 4-14.

Kutcher, S. and Marton, P. (1991) Affective disorders in tirst-de-

gree relatives of adolescent onset bipolars, unipolars, and

normal controls. J. Am. Acad. Child Psychiatry 30, 75-78.

Leslie, R.D.G. and Pyke, D.A. (1980) Identical twins in diabetes.

In W.J. Irvine (Eds.), Immunology of Diabetes. Teviot Scien-

tific Publications, Edinburgh, pp. IOl- 109.

Livingston, R., Nugent, H., Rader, L. and Smith, G.R. (1985)

Family histories of depressed and severely anxious children.

Am. J. Psychiatry 142, 1497-1499.

McGuffin, P. and Katz, R. (1986) Nature, nurture and affective

disorder. In J.F.W. Deakin (Eds.), The Biology of Depression.

Royal College of Psychiatrists, London, pp. 26-52.

McGuffin, P., Katz, R., Aldrich, J. and Bebbington, P. (1988) The

Camberwell Collaborative Depression Study. II. Investigation

of family members. Br. J. Psychiatry 152, 766-774.

Mitchell, J., McCauley, E., Burke, P., Calderon. R. and Schloredt,

K. (1989) Psychopathology in parents of depressed children

and adolescents. J. Am. Acad. Child Adolesc. Psychiatry 28,

352-357. Norusis, M.J. (I 990) SPSS/PC + Manual. SPSS Inc., Chicago,

IL. Numberger, J.I. and Gershon, G.S. (1984) Genetic of affective

disorders. In R.M. Post and J.C. Ballenger (Eds.), Neurobiol-

ogy of Mood Disorder. Williams&Wilkins, Baltimore, MD,

pp. 76-101.

Pearce, J.B. (1978) The recognition of depressive disorder in

children. J. R. Sot. Med. 71, 494-500.

Perris, C. (1992) Bipolar-unipolar distinction. In E.S. Paykel

(Eds.), Handbook of Affective Disorders. Churchill Living-

stone, Edinburgh, pp. 75-75.

Plomin, R. and Bergaman, C.S. (1991) The nature of nurture

genetic influences on environmental measures. Behav. Brain

Sci. 14, 373-427.

Puig-Antich, J., Goetz, D., Davies, M., Kaplan, T., Davies, S.,

Ostrow, L., Asnis, L., Twomey, J., Iyengar, S. and Ryan, N.D.

(1989) A controlled family history study of prepubertal major

depressive disorder. Arch. Gen. Psychiatry 46, 406-418.

Quinton, D. and Rutter, M. (1985) Family pathology and child

psychiatric disorder: a four year prospective study. In A.R.

Nicol (Eds.), Longitudinal Studies in Child Psychology and

Psychiatry. Wiley, Chichester.

Rutter, M. (1985) Family and school influences on behavioural

development. J. Child Psychol. Psychiatry 26, 349-368.

Rutter, M. and Quinton, D. (1984) Parental psychiatric disorder:

effects on children. Psychol. Med. 14, 853-880.

Rutter, M. and Redshaw, J. (1991) Growing up a twin: twin-sin-

gleton differences in psychological development. J. Child Psy-

chol. Psychiatry 32, 885-895.

Rutter, M.,Tizard, J. and Whitmore, K. (1970) Education, Health

and Behaviour. Longmans, London.

Rutter, M., Bolton, P., Hanington, R., Le Couteur, A., Macdon-

ald, H. and Simonoff. E. (1990) Genetic factors in child

psychiatric disorders. I. A review of research strategies. J.

Child Psychol. Psychiatry 3 I, 3-37.

Schwartz, C.E., Darer, D.J., Beardslee. W.R., Lavori, P.W. and

Keller, M.B. (1990) Maternal expressed emotion and parental

affective disorder: risk for childhood depressive disorder, sub-

stance abuse, or conduct disorder. J. Psychiatr. Res. 24, 23 I -

250.

SERC (1990) EGRET User’s Manual. Statistical and Epidemio-

logical Research Corporation, Seattle, WA.

Shepherd, M., Oppenheim, B. and Mitchell, S. (1971) Childhood

Behaviour and Mental Health. University of London Press,

London.

Simonoff, E., McGuffin, P. and Gottesman, 1.1. (1994) Genetic

influences on normal and abnormal development. In M. Rut-

ter, E. Taylor and L. Hersov (Eds.), Child and Adolescent

Psychiatry. Modem Approaches. Blackwell, Oxford, pp. 129-

151.

Software, B.S. (I 990) BMDP Statistical Software Manual, Vol. 2.

University of California Press, Berkeley, CA.

Spitzer, R.L. and Endicott, J. (1975) Schedule for Affective

Disorders and Schizophrenia - Lifetime Version. Biometrics

Research, New York, NY.

Spitzer, R.L., Endicott, J. and Robins, E. (1978) Research Diag-

nostic Criteria: rationale and reliability. Arch. Gen. Psychiatry

35, 773-782.

Strober, M. (1992) Relevance of early age-of-onset in genetic

studies of bipolar affective disorder. J. Am. Acad. Child

Psychiatry 3 1, 606-610.

Strober, M., Lampert, C., Schmidt, S. and Morrell, W. (1993) The course of major depressive disorder in adolescents: I. Recov-

ery and risk of manic switching in a 24month prospective,

naturalistic follow-up of psychotic and nonpsychotic subtypes.

J. Am. Acad. Child Psychiatry 32, 34-42.

Strober, M., Morrell, W., Burroughs, J., Lampert, C., Danforth, H.

and Freeman, R. (I 988) A family study of bipolar I disorder in

adolescence. Early onset of symptoms linked to increased familial loading and lithium resistance. J. Affect. Disord. 15,

255-268.

Thapar, A. and McGuffin, P. (1994) A twin study of depressive

symptoms in childhood. Br. J. Psychiatry 165, 259-265.

Thorley, G. (I 982) The Bethlem Royal and Maudsley Hospitals’

clinical data register for children and adolescents. J. Adolesc.

5, 179-189.

Thorley, G. (1986) Towards a hyperkinetic syndrome. In E. Taylor (Eds.), The Overactive Child. Blackwell Scientific,

Oxford, pp. 161-173.

22 R. Hurringfon et al./Journul of‘Aj&tive Disorders 42 (1997) 9-22

Thorley, G. (1987) Factor study of a psychiatric child rating scale:

based on ratings made by clinicians on child and adolescent

clinic attenders. Br. J. Psychiatry 150, 49-59.

Todd, R.D., Neuman, R., Geller, B., Fox, L.W. and Hickok, J.

(1993) Genetic studies of affective disorders: should we be

starting with childhood onset probands? J. Am. Acad. Child

Adolesc. Psychiatry 32, 1164-l 171.

Weissman, M.M., Gershon, ES., Kidd, K.K., Prusoff, B.A., Leck-

man, J.F., Dibble, E., Hamovit, J., Thompson, W.D., Pauls.

D.L. and Guroff, J.J. (1984aJ Psychiatric disorders in the

relatives of probands with affective disorders. The Yale Uni-

versity-National Institute of Mental Health Collaborative

Study. Arch. Gen. Psychiatry 41. 13-21.

Weissman, M.M., Wickramarame, P., Merikangas, K.R., Lect-

man, J.F., Prusoff, B.A., Caruso, K.A., Kidd, K.K. and Gan-

mon, D. (l984bJ Onset of major depression in early adulthood.

Increased familial loading and specificity. Arch. Gen. Psychia-

try 41. 1136-1143.

Weissman, M.M., Gammon, G.D., John, K., Merikangas, K.R.,

Warner, V., Prusoff, B.A. and Sholomskas, D. (1987) Chil-

dren of depressed parents: increased psychopathology and

early onset of major depression. Arch. Gen. Psychiatry 4.4,

847-853.

Weissman. M.M., Warner, V., Wickramarame, P. and Prusoff,

B.A. (I 988) Early-onset major depression in parents and their

children. J. Affect. Disord. 15, 269-277.

Wolkind, S. and Rutter, M. (1985) Separation, loss and family

relationships. In M. Rutter and L. Hersov (Eds.), Child and

Adolescent Psychiatry: Modem Approaches. Blackwell, Ox-

ford, pp. 34-57.

Zeitlin, H. (1986) The Natural History of Psychiatric Disorder in

Children. Maudsley Monograph 29. Oxford University Press,

Oxford.

Documents

Psychiatric disorders in the relatives of depressed probands I. Comparison of prepubertal, adolescent and early adult onset cases