The influence of an uncaring partner on the type and outcome of depression

Journal of Affective Disorders 66 (2001) 207–214 www.elsevier.com / locate / jad

Research report

The influence of an uncaring partner on the type and outcome of depression Gordon Parker*, Julie Ritch Mood Disorders Unit and School of Psychiatry, University of New South Wales, Prince of Wales Hospital, Randwick, NSW 2031, Sydney, Australia Received 27 April 2000; received in revised form 31 July 2000; accepted 30 August 2000

Abstract Background: We seek to clarify recent inconsistent research findings in relation to the Intimate Bond Measure (IBM), where low IBM care scores have been held to be over-represented in non-melancholic depression and to predict a poor depressive episode outcome. Methods: A sample of 82 subjects meeting DSM criteria for a major depressive episode lasting less than 2 years, took part in a 1-year follow-up study. The IBM was completed at initial assessment and depression severity assessed at baseline and at follow-up assessment, allowing a measure of ‘outcome’. Results: A significantly greater proportion of DSM-defined non-melancholic than melancholic depressives perceived their partner as providing deficient care and were classified as being in a dysfunctional relationship, seemingly unrelated to demographic differences or by depression severity, chronicity or recurrence. IBM scores were again established as independent of a number of putative distorting influences, such as depression severity and ‘neuroticism’. Although the IBM failed to significantly predict outcome, there was a clear trend for greater improvement in those in IBM care score-defined ‘functional relationships’. Limitations: The study failed to assess the change in patients’ perceptions of care over time, which may have contributed to the failure to replicate past findings in relation to outcome. Conclusions: The significance of examining the relevance of psychosocial factors such as deficient intimacy to separate depressive sub-types is highlighted. Implications of these findings for clinical intervention and future research studies are discussed.  2001 Elsevier Science B.V. All rights reserved. Keywords: Depression; Intimate relationships; Intimate Bond Measure; Melancholic and non-melancholic

1. Introduction Social factors have long been implicated as both risk factors to depression and as a predictor of its outcome. Sarason et al. (1997) detailed the role of social support to health outcomes, highlighting the *Corresponding author. Fax: 1 61-2-9382-4343.

significance of the degree to which a person perceives support is available, as well as characteristics of the relationship between the supporter and the recipient. They observed that global measures of social support (as generally used in the field) are often relatively weak predictors of outcome, usually due to distortion by confounding variables. They argued that, since the most meaningful social contact

0165-0327 / 01 / $ – see front matter  2001 Elsevier Science B.V. All rights reserved. PII: S0165-0327( 00 )00311-6

208

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

for many people is provided by their intimates, specific information about the qualities of close relationships significantly contributes to our understanding of a variety of health issues, including depression An association between deficient intimacy and an increased risk of depression has been identified in a number of studies (e.g., Brown and Harris, 1978; Costello, 1982; Harris and Brown, 1985; Weissman, 1987; Rodriguez et al., 1993). Deficient intimacy has also been viewed as a predictor of the course and outcome of depression (e.g., Surtees et al., 1983; Lin et al., 1986; Frank et al., 1989; George et al., 1989; Hooley and Phil, 1990), including studies using the ‘expressed emotion’ paradigm, and where the level of criticism expressed by a spouse at the time of hospital admission predicted symptomatic relapse for depressed patients (Vaughn and Leff, 1976). As noted by Waring (1985), many of the measures used to assess aspects of interpersonal relationships are flawed by failure to isolate personality characteristics of the patient and by distorting effects of current mood state from measurement of the actual interpersonal relationship. Wilhelm and Parker (1988) developed the Intimate Bond Measure (IBM) to measure key elements of an intimate relationship, and tested the measure for any such limitations. The IBM is a 24-item self-report questionnaire which measures a respondent’s perception of their partner’s attitudes and behaviours toward them, and has both ‘care’ and ‘control’ scales. The care scale, the focus of this study, has items assessing care expressed emotionally as well as physically, with constructs of warmth, consideration, affection and companionship. IBM care scale scores have been shown to correlate with ‘care’ ratings generated by experienced clinicians and marital therapists (Wilhelm and Parker, 1988), arguing that IBM care scores reflect both ‘actual’ and ‘perceived’ levels of care. As noted earlier, any ‘intimacy’ measure is limited in studying depressed samples if measure scores are influenced by depressed mood or neurotic personality traits. The IBM ‘care’ scale — but not the ‘control’ scale — has been shown to be relatively free of distortions produced by state depression (Wilhelm and Parker, 1988; Hickie et al., 1991). Studies have revealed that it also appears uninfluenced by recurrence of depres-

sion, socio-demographic factors and neuroticism, and that scores are relatively stable over time (Hickie et al., 1990, 1991). Initial studies using the IBM measure confirmed past conclusions about the risk effected by deficient intimacy on increasing the risk of post-natal depression (Boyce et al., 1991) and of depressive disorder (Hickie et al., 1990, 1991), but results in the latter studies varied across depressive sub-types. In one report (Hickie et al., 1990), deficient care was two to three times more likely to be reported by patients with non-endogenous (or non-melancholic) depression than matched non-depressed subjects but no more likely to be reported by patients with endogenous or melancholic depression, and thus compatible with the ‘binary view’ of depression (i.e., that certain psychosocial factors are specifically over-represented in non-melancholic’ depression). If interpersonal relationships are of key relevance, then ‘deficient’ intimacy should also have an impact on the outcome of depression. In one study of a non-melancholic sample, IBM ‘care’ scale scores accounted for the major proportion of the variance in depressive outcome (Hickie and Parker, 1992). A New Zealand study, however, provided only partial replication of these findings. Mulder et al. (1996) studied 105 depressed patients, and, while IBM ‘care’ scores were independent of depression severity and personality measure scores, they did not differ across melancholic and non-melancholic depressive sub-samples, and did not predict treatment outcome. When their sample was restricted to those previously treated with antidepressants, those with ‘recurrent’ melancholic depression had IBM care scores in the ‘normal’ range, while those with ‘recurrent’ non-melancholic depression had low care scores. The authors therefore concluded that the differential care scores might reflect the impact of recurrent episodes of depression on interpersonal relationships in non-melancholic depression, and that individuals with melancholic depression may respond more quickly to treatment — so preventing depressive episodes having distinct impact on their marital relationship. They also speculated whether their failure to replicate earlier study results might reflect sample selection differences, with their sample having a large proportion of out-patients, and

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

presumably not weighted to treatment-resistant subjects. Burns et al. (1994) have also queried whether previous investigators may have overestimated the aetiological impact of interpersonal relationship components on depression, suggesting that outcome differences may emerge more from the impact of issues such as previous treatment and depression severity. Such inconsistent findings lead to the current independent study. Specifically, we sought to: (a) determine whether IBM scores differ across defined melancholic and non-melancholic depressive patients; (b) determine the extent to which IBM care scores are associated with socio-demographic factors, as well as with severity, recurrence and past treatment of depression and neuroticism, and if so, whether these variables account for any differences between melancholic and non-melancholic sub-sets; and, finally, (c) to identify whether IBM care scores predicted depression outcome.

2. Methods

2.1. Subjects and study design The initial sample comprised both consecutive depressed in-patients and out-patients of our Mood Disorders Unit (MDU), as well as routine out-patients of several other Sydney hospitals, but did not include subjects studies in our earlier samples. All subjects met DSM-III-R (American Psychiatric Association, 1987) criteria for a major depressive episode present for less than 2 years (to redress excessive weighting to chronicity or treatment resistance) and did not have a higher order diagnosis (i.e., schizophrenia, alcoholism or dementia), although those with co-morbid conditions such as anxiety and personality disorders were included. Of the 270 patients recruited from 1993 to 1996, a sub-group of 82 subjects satisfied criteria for this study (i.e., being in an intimate relationship at the time of study recruitment and completing an IBM for their partner), while 32 subjects who nominated being in an intimate relationship failed to complete the IBM measure.

209

2.2. Defining ongoing relationships Relevant subjects were those in a current relationship that had existed for more than 12 months (i.e., either married, engaged to be married, or in an ongoing sexual relationship), and who viewed that relationship as their principal interpersonal relationship. At baseline they completed the IBM (Wilhelm and Parker, 1988) in relation to that partner. Subjects were categorised as having a ‘dysfunctional’ relationship according to a previously defined cut-off IBM care score of less than 20 (Hickie and Parker, 1992), while IBM scores were also examined dimensionally in some analyses.

2.3. Assessment of depression variables Depression severity was assessed by the self-report Beck measure or BDI (Beck et al., 1961), both at baseline and at the 12-month follow-up. All patients were assigned as either having a melancholic (‘mel’) or non-melancholic (‘non-mel’) depressive sub-type at baseline by applying ‘DSM-IV’ (American Psychiatric Association, 1994) decision rules, while the interviewing psychiatrist was also required to rate whether patients met DSM-IV ‘recurrent’ episode criteria. A research psychologist undertook a semi-structured interview at the baseline assessment, assessing details about any past depressive episodes and past treatment, and sought extensive details about current symptoms. At a 12-month review, the BDI was readministered to assess the level of 12-month improvement in depression severity while we also assessed whether or not subjects then met DSM-IV criteria for a major depressive episode.

2.4. Other relevant baseline data Socio-demographic information was collected at the baseline assessment, with social class determined by a seven-point occupation-based scale. The interviewing psychiatrist was required to undertake a clinical assessment of each subject’s personality style. In this report, we focus on any effect of neurotic personality traits on IBM scores, and thus note that neuroticism or ‘anxious personality style’

210

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

(as assessed by the psychiatrist) was here defined as ‘‘nervy, tense and a worrier’’ and scored on a sixpoint scale ranging from 0 to 5.

2.5. Statistical analyses Independent t-tests were used for dimensional variables and x 2 analyses for categorical data — while the odds ratio statistic was used to approximate the size of any association between the categorical variables. Pearson correlational analyses were used to test associations between variables. Logistic and multiple regression analyses were performed to determine whether differences between the melancholic and non-melancholic groups remained significant whilst controlling for several factors, and to determine the contribution of various predictor variables to the outcome of depression as measured by BDI percentage improvement scores.

3. Results

3.1. Description of the sample The sample of 82 subjects had an average age of 43.3 (S.D. 12.6; range, 21–77) years; 34 (42%) were males; and 65% were in-patients. Seventeen (21%) had never previously had psychiatric contact, and 20 (24%) had never had a previous depressive episode. The mean IBM ‘care’ score in the sample was 26.5 (S.D. 5 8.6). The mean baseline BDI score was 27.6

(S.D. 5 11.5), while the follow-up BDI score was 15.2 (S.D. 5 12.9). The DSM-IV system classified 34 patients (41.5%) as ‘melancholic’ (‘mels’). Table 1 details principal characteristics of the mel and non-mel samples, with no differences identified on socio-demographic variables, out / in-patient status, lifetime episodes of depression, baseline and follow-up BDI scores and the percentages rating positive on the neuroticism measure. This was a beneficial result in suggesting that these variables would be then unlikely to confound principal analyses.

3.2. Associations between socio-demographic variables, in-patient status, severity of depression, neuroticism and the IBM scale In order to examine possible distorting influences on comparisons of the mel and non-mel subjects, IBM care scores were examined against several key variables within the whole sample. Care scores were not associated with age (r 5 0.07) or social class (r 5 2 0.05). There were trends for care scores to be associated with sex (mean male score for partner 5 28.6; mean female score for partner 5 25.0, t 5 1.9, P 5 0.06), but not with in-patient / out-patient status (i.e., 26.6 vs. 26.5). IBM care scores were independent of baseline depression severity (as assessed by the BDI) in the whole sample (r 5 2 0.02). However, this overall coefficient was the sum of a weak positive coefficient (r 5 0.16, P 5 0.52) in the nonmel subjects, and a negative correlation (r 5 2 0.34,

Table 1 Differences between DSM-IV-assigned non-melancholic and melancholic depressive subjects on descriptive variables

Sex (males, %) Age Socio-economic status Out-patients (%) Lifetime episodes of depression Initial depression severity (BDI score) Follow-up depression severity (BDI score) ‘‘Nervy tense and a worrier’’ (neuroticism) (%)

‘Non-melancholic’ subjects

‘Melancholic’ subjects

Test

P

35 42.1 (S.D. 5 12.17) 4.4 71 3.1

50 44.8 (S.D. 5 13.26) 4.2 56 3.3

x 2 5 1.74 t 5 0.9 t 5 0.6 x 2 5 1.9 t 5 0.2

0.19 0.35 0.57 0.16 0.83

27.2

28.2

t 5 0.4

0.69

16.1

13.9

t 5 0.7

0.50

60.4

58.8

x 2 5 0.02

0.89

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

P 5 0.07) in the mels. IBM scores were independent of psychiatrist-rated neuroticism scores in the whole sample (r 5 2 0.13) and, separately, in the non-mels (r 5 0.09) and mels (r 5 0.21).

211

Table 2 Differences between IBM ‘care’ scores as a function of recurrent depression and previous episodes and treatment in the DSM-IVdefined non-melancholic sample t-test

P

DSM-IV-defined ‘recurrence’ Recurrent (n 5 29) 22.8 Non-recurrent (n 5 16) 25.9

1.1

0.26

The non-mels returned lower IBM care scores for their partners than did the mels (i.e., 30.3 vs. 23.8, t 5 3.60, P , 0.001). Using the pre-established IBM cut-off score of 20 or less as indicative of ‘low care’, then significantly more of the non-mels than the mels (33% vs. 12%, x 2 5 5.02, P , 0.02) so scored their partner, with the odds ratio being 3.75.

Any previous episode Yes (n 5 37) No (n 5 11)

0.1

0.90

0.9

0.40

3.4. Associations between IBM scores and recurrent depression and previous treatment in the non-melancholic subjects

controlling for (i) ‘recurrent episode’. (ii) ‘previous episode’ status, (iii) any previous antidepressant treatment, (iv) sex, (v) baseline BDI-defined depression severity and (vi) neuroticism scores. The logistic regression failed to identify a significant contribution from any of those ‘control’ variables, with IBM care scores being the only significant variable (P 5 0.005) in the equation, establishing that IBM care scores did differ across the diagnostic sub-types after accounting for such potential confounding variables.

3.3. The differential likelihood of dysfunctional intimate relationships in non-melancholic and melancholic depression

DSM-IV criteria for recurrence were met by 61% of the mels and 64% of the non-mels; 65% of the mels and 77% of the non-mels had had a previous episode; and 87% of the mels and 85% of the non-mels had previously had antidepressant therapy, with none of these differences being significant. In the introduction we noted the suggestion (Mulder et al., 1996) that recurrence of depression rather than deficient care in an intimate relationship accounts for low IBM ‘care’ scores in the non-mels and so accounts for the differences between the mel and non-mel sub-samples. We examined the impact of recurrence in several ways in the non-mels. Table 2 data reveal no significant differences in IBM scores between the DSM-IV ‘recurrent’ and ‘non-recurrent’ depressive sub-sets, or between those experiencing a ‘previous episode’ or not, or between sub-sets of those who had previously received or not received antidepressant medication.

3.5. Multivariate analysis of IBM care score differentiation Despite the absence of any suggestion from univariate analyses that confounding had occurred, we examined whether IBM care scores differed across melancholic and non-melancholic sub-samples after

Defined sub-samples

Mean IBM care score

23.9 23.5

Any previous antidepressant therapy Yes (n 5 42) 24.2 No (n 5 6) 21.0

3.6. Predictive utility of the IBM In order to test whether the baseline IBM care score predicted improvement in depression, we undertook a number of analyses. An initial analysis included a product term to test the possibility that the association between care scores and BDI follow-up scores was different for the melancholic / non-melancholic groups, but the interaction was not significant (B 5 0.13, t 5 0.31, P 5 0.76). A multiple regression analysis was then undertaken with the follow-up BDI score as the dependent variable and the baseline BDI score and the IBM care score as predictor variables. Only 60 subjects had BDI follow-up scores and were thus able to be included in this analysis. The IBM care score failed to predict BDI outcome scores (B 5 2 1.10, t 5 2 0.31, P 5 0.75). We then focussed on any prediction of outcome in the non-melancholic sample. There were no statistically significant outcome differences between those

212

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

in pre-defined ‘dysfunctional’ (i.e., a care score of less than 20) and ‘functional’ relationships. There were, however, consistent trends in the data, with those in ‘functional’ relationships tending to show higher rates of improvement. Thus, 55% of those in a ‘functional’ relationship improved compared to only 36% of those in a ‘dysfunctional’ relationship — when ‘outcome’ was quantified as a 50% or greater improvement in subjects’ BDI scores. Secondly, 63% in a ‘functional’ relationship, as against 37% in a ‘dysfunctional’ relationship no longer met DSM-IV ‘MDE’ criteria.

4. Discussion This study aimed to clarify the significance of deficient intimate care in having any specificity to non-melancholic depression and as a predictor of depression outcome. As noted earlier, initial Australian studies reported that non-melancholic (compared to melancholic) depressed subjects reported significantly less care from their partners (Hickie et al., 1990), while lower IBM care scores were shown to predict outcome in those with non-melancholic depression (Hickie and Parker, 1992). In the New Zealand study, Mulder et al. (1996) reported that IBM care scores did not differentiate between melancholic and non-melancholic subjects, and did not predict outcome of the depression, and judged that the earlier results may have been confounded by recurrence of depression. In this independent study, we sought to examine whether confounding by recurrence — as well as a range of other theoretically confounding variables (i.e., sex, depression severity, and neuroticism) — may have contributed to any differences between care scores returned by the melancholic and nonmelancholic subjects. In the raw analyses, the nonmelancholic subjects were significantly more likely to return low care scores and were nearly four times more likely to return pre-established ‘low care’ scores, while our multivariate analysis established that diagnostic specificity remained significant after controlling for the putative confounding variables. Such findings support our earlier studies and argue that, in pursuing any risk factor to depression, it is important to examine for any specificity to depres-

sive sub-types. Our results support Matussek and Wiegand’s (1985) findings, which emphasised the selective importance of dysfunctional relationships to non-melancholic depression. In their study, they argued that ‘non-endogenous’ depressives become depressed more often because of partnership-related events than ‘endogenous depressives’. Based on their clinical observation, they also speculated that nonmelancholic depressives choose their partners in a more autonomously purposeful and conscious manner than melancholic depressives who get involved without as much awareness of the situation. In essence, their theory implies that personality differences explain why non-melancholic depressives are more vulnerable and are at greater risk of disappointment and depression in this situation, and so perceive their intimate relationships differently to those with melancholic depression. The antecedents of that vulnerability may also lie in early dysfunctional relationships. In one earlier study (Hickie et al., 1990), we established that those reporting markedly deficient parental care in childhood were highly likely to report very poor current intimate relationships, which, if not a response bias, can be interpreted as reflecting poor parenting laying down a diathesis to later uncaring intimate relationships (and which might occur in multiple ways). In addition, Brown et al. (1994) established a link between poor current interpersonal relationships and having been abused or neglected in childhood. If such associations exist, then intimate care levels may not, of course, have a direct influence on depression levels but be secondary to the effects of childhood adversity. Alternatively, both childhood adversity and current interpersonal stressors may contribute independently, whether iteratively or sequentially, and increase the chance of depression. Differences across the depressive sub-types might also reflect those with melancholic depression viewing their illness as more ‘biological’ and thus being less likely to attribute social factors to episode onset and to be less inclined to blame their partners. We failed to identify significant differences in outcome between those who were in IBM care scoredefined dysfunctional relationships in the whole sample. When we examined those with non-melancholic depression only, there were trends for those in functional relationships to show superior ‘improve-

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

ment’, in being approximately twice more likely to show a significant reduction in Beck depression scores and to no longer meet major depression criteria. Thus, our results were not as clear cut as the previous study (Hickie and Parker, 1992) where the IBM ‘care’ score accounted for a large proportion of the variance in predicting outcome and where even fewer subjects were studied. In that study, IBM scores were most discriminating for short-term reduction in depression (i.e., at 6 weeks) and it may be that we failed to establish a significant difference in this study by not examining prediction of short-term improvement. Study differences may, of course, reflect other factors impacting on the course of depression, including the use and efficacy of antidepressant treatments. Additionally, we did not reassess patients’ intimate relationships at the followup assessment (as in the previous study). Thus, some patients may have separated from uncaring partners, while for others, deficiencies in care may have improved over time, thus influencing the ‘risk factor’. Why results from our studies differ from the New Zealand study (Mulder et al., 1996) remains unclear, although those authors suggested that sample selection may be influential. The Australian studies have been undertaken on samples seemingly over-represented by those with lengthier and treatment-resistant disorders, as well as having a significant percentage of in-patients, while only 4% of the New Zealand sample were in-patients. Further studies should then pursue the extent to which sample characteristics influence results, and the meaning of those sample nuances to the issues under investigation. Our findings nevertheless again highlight the importance of addressing the interpersonal environment and, in particular, deficient care in intimate relationships, and particularly in the non-melancholic depressive disorders. Angst (1999) has recently argued that a conceptual shift has occurred in our understanding of depression, with it now being more viewed as a chronic and recurrent disorder, thus making it essential to identify predictors of chronicity and recurrence — with one such predictor being relationship difficulties. This conceptual shift offers another argument for evaluating components of the depressed patient’s social support system (such as care from an intimate) and their significance in

213

relation to the development and outcome of depression.

Acknowledgements The assistance of Professor Ron Rapee and of Kerrie Eyers, Heather Brotchie, Lana Judelman, David Ritch, Sabine Roussos, Kay Roy and Alan Taylor is noted with appreciation. This study was supported by NHMRC Program Grant (993208) and an Infrastructure Grant from the NSW Department of Health.

References American Psychiatric Association, 1987. DSM-III-R: Diagnostic and Statistical Manual of Mental Disorders, 3rd Edition, revised. American Psychiatric Press, Washington, DC. American Psychiatric Association, 1994. DSM-IV: Diagnostic and Statistical Manual of Mental Disorders, 4th Edition. American Psychiatric Press, Washington, DC. Angst, J., 1999. Major depression in 1998: Are we providing optimal therapy. J. Clin. Psychiatry 60, 5–9. Beck, A.T., Ward, C.H., Mendelson, M., Mock, J., Erbaugh, J., 1961. An inventory for measuring depression. Arch. Gen. Psychiatry 4, 561–571. Boyce, P., Hickie, I., Parker, G., 1991. Parents, partners or personality? Risk factors for post-natal depression. J. Affect. Disord. 21, 245–255. Brown, G.W., Harris, T., 1978. The Social Origins of Depression. Tavistock, London. Brown, G.W., Harris, T.O., Hepworth, C., Robinson, R., 1994. Clinical and psychological origins of chronic depressive episodes II: a patient enquiry. Br. J. Psychiatry 165, 457–465. Burns, D.D., Sayers, S.L., Moras, K., 1994. Intimate relationships and depression: is there a casual connection? J. Consult. Clin. Psychol. 62, 1033–1043. Costello, C.G., 1982. Social factors associated with depression: a retrospective community study. Psychol. Med. 12, 329–339. Frank, E., Kupfer, D.J., Perel, J.M., 1989. Early recurrence in unipolar depression. Arch. Gen. Psychiatry 46, 397–400. George, L.K., Blazer, D.G., Hughes, D.C., Fowler, N., 1989. Social support and the outcome of major depression. Br. J. Psychiatry 154, 478–485. Harris, T., Brown, G.W., 1985. Interpreting data in an etiological studies of affective disorder: some pitfalls and ambiguities. Br. J. Psychiatry 147, 5–15. Hickie, I., Wilhelm, K., Parker, G., Boyce, P., Hadzi-Pavlovic, D., Brodaty, H., Mitchell, P., 1990. Perceived dysfunctional relationships; a specific association with the non melancholic depressive subtype. J. Affect. Disord. 19, 99–107. Hickie, I., Parker, G., Wilhelm, K., Tennant, C., 1991. Perceived

214

G. Parker, J. Ritch / Journal of Affective Disorders 66 (2001) 207 – 214

interpersonal risk factors to non-endogenous depression. Psychol. Med. 21, 399–412. Hickie, I., Parker, G., 1992. The impact of an uncaring partner on improvement in non-melancholic depression. J. Affect. Disord. 25, 147–160. Hooley, J.M., Phil, D., 1990. Expressed emotion and depression. In: Keitner, G.I. (Ed.), Depression and Families: Impact and Treatment. American Psychiatric Press, Washington, DC, pp. 55–84. Lin, N., Dean, A., Ensel, W.M., 1986. Social Support, Life Events, and Depression. Academic Press, San Diego, CA. Matussek, P., Weigand, M., 1985. Partnership problems as causes of endogenous and neurotic depression. Acta. Psychiatr. Scand. 71, 95–104. Mulder, R.T., Joyce, P.R., Sullivan, P.F., Oakley-Browne, M.A., 1996. Intimate bonds in depression. J, Affect. Disord. 40, 175–178. Rodriguez, V.B., Bayon, C., Franco, B., Canas, F., Graell, M., Salvador, M., 1993. Parental rearing and intimate relations in women’s depression. Acta. Psychiatr. Scand. 88, 193–197. Sarason, B.R., Sarason, I.G., Gurung-Regan, A.R., 1997. Close

personal relationships and health outcomes: a key to the role of social support. In: Duck, S. et al. (Ed.), Handbook of Personal Relationships: Theory, Research and Interventions, 2nd Edition. Wiley, Chichester, UK. Surtees, P.G., Dean, C., Ingham, J.G., Kreitman, N.B., Miller, P.M., Sashidharan, S.P., 1983. Psychiatric disorder in women from an Edinburgh community: associations with demographic factors. Br. J. Psychiatry 142, 238–246. Vaughn, C.E., Leff, J.P., 1976. The influence of family and social factors on the course of psychiatric illness: a comparison of schizophrenic and depressed neurotic patients. Br. J. Psychiatry 129, 125–137. Waring, E.M., 1985. Measurement of intimacy: conceptual and methodological issues of studying close relationships. Psychol. Med. 15, 9–14. Weissman, M.M., 1987. Advances in psychiatric epidemiology: rates and risks for major depression. Am. J. Public Health 77, 445–451. Wilhelm, K., Parker, G., 1988. The development of a measure of intimate bonds. Psychol. Med. 18, 225–234.