In a sample of 730 women in their third trimester of pregnancy recruited in and around Ankara, we found that lower quality ratings for three key relationships - with the spouse, mother and mother-in-law - were associated with case level depression as defined from the Edinburgh Postnatal Depression Scale (EPDS). The association between lower quality spouse relationship and depression was stronger in women living in a traditional rather than nuclear family arrangement.
As discussed earlier, antenatal depression persists into the postnatal period in a large proportion of cases and many cases of postnatal depression begin in the antenatal period [6
]. Antenatal depression therefore represents an important clinical and public health issue [8
] because of the potential for early intervention. Although there have been reports of puerperal psychosis preceded by antenatal depression [32
], postnatal depression as the much more common outcome is likely to be the most important focus for prevention. However, as discussed, there has been relatively little research into antenatal depression and its correlates. Findings from this study show similarities and dissimilarities compared to other studies in terms of potential risk factors for antenatal depression. Social support, life events, violence were associated with depression in one study whereas age, education level and income were not [33
]. In another study, risk factors for depression during pregnancy included younger age and lower education [35
]. Increased parity and lack of support, particularly poor support from the partner/husband have also been associated with depression in both developed and developing countries [34
]. In particular, physical abuse by intimate partners before or during pregnancy has been found to be a particularly important potential risk factor for antenatal depression [35
]. In this analysis, the focus was on social support and relationship quality specifically; however, unadjusted analyses revealed associations with increased parity, worse self-rated physical health, higher number of life events and self-reported past emotional problems. On the other hand, depression was not significantly associated with age, education, income or traditional/nuclear family structure. These findings suggest at least some level of heterogeneity between settings in correlates of antenatal depression, although methodological differences with respect to sampling and measurement cannot be excluded as an underlying reason for this.
Although it was not the primary objective of this analysis to investigate differences in the prevalence of antenatal depression between traditional and nuclear family settings, the observed lack of difference is potentially interesting. In the Turkish context, the two family models have co-existed for many decades and there is relatively little stigma attached to women living in either family model. In particular, we do not feel that women in nuclear settings have had to 'extract' themselves from traditional settings and thus we do not feel that the woman's personality or the attitudes of her family are likely to be a major factor. Largely, the family model in which a woman is living depends on issues such as the availability of work and accommodation. Further analysis would be required to clarify whether there was any negative confounding, obscuring a true difference in depression prevalence between settings. However, this was beyond the scope of this paper which sought to focus on associations between depression and social relationships and the role of family structure as an effect modifier rather than as an exposure itself.
To our knowledge, ours is the first study which has assessed the association between antenatal depression and support from the mother and mother in-law, although this has been investigated previously for postnatal depression in Turkey [36
]. In our sample we found strong associations between depression and nearly all measures of social support from the three relatives in question. Those with the husband and mother-in-law were particularly strong, which is consistent with the importance of these figures in women's lives in this culture. The only exception was that negative aspects of the relationship with the mother were more strongly associated with depression than those with the mother-in-law. However, this might possibly reflect a long-standing poor parental relationship prior to marriage but with lasting effects on mental health.
As mentioned earlier, Turkey in general (and Ankara in particular) offers important advantages for research into the role of different family structures because of the longstanding co-existence of 'Western' and traditional 'Middle Eastern' cultures. The relationship between women and their mother and mother in-law is still important in Turkish culture, whether the woman is living in a nuclear or extended family setting. In Turkish traditional settings, a woman will typically move to live with her husband and his family in the same house when she gets married. In this setting, the expected role of a woman's own mother is to support this marriage by helping her daughter on practical issues (e.g. taking care of children) and emotional issues. The study was specifically set up to investigate these issues, funded through the Wellcome Trust's 'Health Consequences of Population Change' programme which sought to support research into the potential health impacts of rapidly changing societies. We investigated whether an extended family setting might modify potential effects of spousal and other key relationships on depression risk, specifically hypothesising that the presence of other family members would reduce the impact of a poor quality spousal relationship. Contrary to our hypothesis, effect modification in the opposite direction was found with stronger associations between spousal support and depression in traditional families, particularly with respect to lower emotional support as an exposure. This requires confirmation in other samples. However, it may reflect a higher visibility of marital difficulties in extended families and hence a stronger impact on depression. It might also reflect families taking the side of the husband and feelings of isolation of the woman in question. The stronger association in women without previous children might reflect a buffering effect of other children on the impact of marital strain or possibly higher feelings of empowerment in this group of women and/or the presence of children allowing greater access to friends and extra-familial support networks.
Also of interest was the observation that the association with social support from the mother-in-law was equally strong in nuclear and traditional families, emphasising the importance of this relationship in Turkish culture, and with implications for future clinical and public health interventions. The association with support from the mother was, as mentioned, weaker in most respects, and the observation of possibly opposite associations with daughters' depression between traditional and nuclear families might reflect differing roles of the mother in the two situations. Higher practical support from the mother in the context of an extended family structure (i.e. for women living with their husband's family) might represent a more severe breakdown of relationships in the household where women are residing. Support from family members has been found to be an important buffer against depression in women from other low and middle income settings [37
]. Some research into perinatal mental disorder in Islamic nation settings has suggested both high prevalence of disorder and a potentially harmful role of disruptions to traditional family structures [38
]. Although a high prevalence of antenatal depression was found in our sample, consistent with this, there was little evidence that traditional family structures conferred additional protection, either directly or through buffering effects of individual relationships. However, it should be borne in mind that these nuclear and traditional structures have co-existed in Turkey for a long time, potentially allowing individual and societal adjustment. Results cannot necessarily be generalised to nations or cultures undergoing more rapid changes and further research is required in these settings.
Strengths of this study include the particular features of the setting, as mentioned, the large and heterogeneous sample, the standardised assessment instruments which have been well-validated in a variety of international settings, and a comprehensive range of covariates. Random sampling of antenatal clinics was not feasible in this setting because of difficulties in enumeration of these. An approach was taken instead to maximise the heterogeneity of populations served which we believe constituted the next best approach to sampling. Response rates were relatively high and we believe that the findings should generalise to the source populations. The Edinburgh Postnatal Depression Scale used in this study is, as stated, widely used in international research. However, it should be borne in mind that it is a screening instrument, measuring number of depressive symptoms and does not seek to define specific depression syndromes. Furthermore, it is possible that other syndromes such as anxiety may influence caseness on this instrument. 'Depression' is therefore used as a shorthand term to describe case level symptomatology on this instrument, but it should be borne in mind that this is not synonymous with a clinical diagnosis and that clinical. Other principal limitations arise from the cross-sectional nature of this analysis. In particular, associations between lower social support and depression might reflect response bias in people with depression, or might reflect an effect of depression (the current episode and/or earlier episodes) on interpersonal relationships and actual levels of support, as well as the causal relationship of interest between low social support and risk of depression. In this respect, a key limitation is that there was little information on history of previous depressive episodes, whether known or unknown to clinical services, and further follow-up of this sample is currently underway which will seek to address these issues. Confounding factors were addressed as comprehensively as possible; however, residual confounding cannot be excluded. For example, personality was not measured although this is a factor that could have potentially influenced interpersonal relationships as well as risk of depressive episodes.