Examining whether offspring psychopathology influences illness course in mothers with recurrent depression using a high-risk longitudinal sample.

Procedure

Data were collected from parents and children via semistructured diagnostic interviews. The study was approved by Multicenter Research Ethics Committee. Families were interviewed at their home (99%) or at the University Hospital of Wales. Written informed consent or assent was obtained from all participants as appropriate. Further information regarding the procedures is detailed elsewhere ( Mars et al., 2012 ).

Measures

Child symptoms of major depressive disorder (MDD) and DBDs. Child symptoms of MDD and DBD were assessed using a semistructured interview, the Child and Adolescent Psychiatric Assessment ( Angold & Costello, 2000 ). Child reports were used to examine the presence of symptoms of depression and DBD (conduct disorder/oppositional defiant disorder) in the preceding 3 months. The total number of child DSM–IV symptoms of depression and DBD, at both baseline and at follow-up, were summed to create depression and DBD symptom counts at both time points. Descriptive data on child symptoms and rates of disorder at baseline and follow-up are presented in Table 1 . Interrater reliability was high (average κ = .94).

Maternal depression symptom levels. The number of maternal DSM–IV depression symptoms was assessed using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN; Wing et al., 1990 ), a semistructured clinical interview assessing depression symptoms in the preceding month. Symptom counts ranged from 0–9. Descriptives of maternal symptoms at baseline and follow-up are presented in Table 1 .

Between the baseline and follow-up assessment, additional information was collected regarding maternal DSM–IV MDD using the SCAN and a life-history calendar approach ( Belli, 1998 ; Caspi et al., 1996 ). The presence of DSM–IV MDD occurring between the baseline assessment and follow-up was established using this data to derive a measure of MDD recurrence. Depression recurrence was considered as a binary variable; where mothers reported MDD at any point between the baseline assessment and follow-up, recurrent depression was considered to be present.

Preliminary Analyses and Missing Data

All analyses adjusted for past maternal depression severity. Interviews at baseline ascertained parents’ periods of hospitalization for depression and impairment of the worst two episodes of depression using Global Assessment of Functioning (GAF) scores ( American Psychological Association, 1994 ). Past maternal depression severity was defined as a GAF score <30 or any hospitalization due to depression in accordance with previous criteria ( Hammen & Brennan, 2003 ). Furthermore, this measure of maternal severity has previously been shown to be associated with negative child outcomes in the current sample ( Mars et al., 2012 ). Past maternal severity was coded as a dichotomous variable (yes/no), with 28.6% reporting at least one severe episode of depression. There was weak evidence of an association between maternal past depression severity and baseline maternal depression symptoms (boy model: b = .31, 95% CI [−.01, .62], p = .057; girl model: b = .26, 95% CI [−.01, .53], p = .057). There was evidence of an association between maternal past depression severity and baseline adolescent depression symptoms for girls (boy model: b = .14, 95% CI [−.07, .36], p = .197; girl model: b = .22, 95% CI [.01, .43], p = .041), but not between maternal past depression severity and baseline adolescent DBD symptoms (boy model: b = −.08, 95% CI [−.36, .20], p = .561; girl model: b = .19, 95% CI [−.03, .40], p = .085).

Given the large age range of the children in the current study (age 9–17 years at baseline), the current study also adjusted for child age. Child age at baseline was associated with girls’ symptoms of depression (girls b = .08, 95% CI [.04, .13], p = .001; boys b = .02, 95% CI [−.03, .07], p = .369). Child age at baseline was also associated with child symptoms of DBD (girls b = .12, 95% CI [.08, .16], p = .001; boys b = .08, 95% CI [.01, .15], p = .018). Both past maternal depression severity and child age were therefore included in the models, but not presented for ease of interpretation.

Of the eligible sample ( n = 299), 14 families refused to participate at follow-up (4.7%). Of the remaining sample ( n = 285), 72 ( n = 25%) had partially complete data (36 families completed questionnaire assessments only and therefore did not have interview data. An additional 36 families had incomplete interview assessments, and therefore, symptom counts could not be computed). Therefore, complete data at baseline and follow-up was available for 73% ( n = 213/299) of the sample.

Families who had missing data at follow-up included mothers with higher depression symptoms at baseline (missing at follow-up M = 3.44, SD = 2.74; participated at follow-up M = 2.38, SD = 2.58, p = .004). There was weak evidence of an association between missing data at follow-up and offspring depression symptoms (missing at follow-up M = 1.48, SD = 1.87; participated at follow-up M = 1.04, SD = 1.51, p = .057). Child DBD symptoms at baseline were not associated with missingness at follow-up (missing at follow-up M = 3.47, SD = 2.41; participated at follow-up M = 2.90, SD = 2.41, p = .522). Missing data at follow-up was also found to vary by maternal education ( p = <.001) and child IQ ( p = <.001). To account for observed patterns of selective attrition, missing data for offspring symptomatology, maternal depression, and other covariates were imputed using multivariate imputation by chained equations ( Van Buuren & Oudshoom, 2000 ) using all available data. This assumes that data are missing at random (MAR) that is, given the observed data included in the imputation model, the missingness mechanism does not depend on the unobserved data ( White, Royston, & Wood, 2011 ). The variables associated with nonresponse were therefore included in the imputation model to make the assumption of MAR as plausible as possible, along with other measures closely associated with maternal and offspring psychopathology (child emotional and conduct symptoms and maternal depression symptoms assessed using questionnaire screening measures at multiple time points) and all other variables included in analyses ( White et al., 2011 ).

Imputation models were run using binary and ordinal logistic and linear regression models as appropriate. Predictive mean matching (PMM) was used when continuous variables were not normally distributed. PMM yields acceptable estimates while maintaining the underlying distribution of the data ( Vink, Frank, Pannekoek, & van Buuren, 2014 ). Where gender interactions were tested, imputation models were run by child gender. All variables with missing data used in analyses were imputed up to the maximum sample size of n = 299. Thirty imputed datasets were derived each with 10 cycles of regression switching and then all analyses were run on imputed datasets by combining estimates using Rubin’s rules ( White et al., 2011 ). All analyses presented show the results using the imputed sample of 299. Complete case analyses were also conducted and are shown in supplementary material.

Analysis

Initial binary logistic regression analyses examined the associations between child depression and DBD symptoms at baseline and the presence of maternal depressive episode recurrence after baseline (i.e., throughout the duration of the study period yes/no). Unadjusted models, and analyses controlling for past maternal depression severity and household income, are reported using risk ratios and 95% confidence intervals. Analyses were conducted in STATA version 13 ( StataCorp, 2007 ).

Cross-lagged and reciprocal effects models utilize longitudinal data to test the direction of effects between variables ( Kline, 2005 ). The cross-lagged model involved simultaneously estimating the contribution of each variable at baseline (maternal depression symptoms and child psychopathology) in accounting for each of the variables at follow-up (maternal depression symptoms and child psychopathology) while controlling for the previous correlation between the two constructs at baseline and the stability in the constructs over time. Reciprocal effects models are used to test the direction of effects that exist within time rather than across time, again controlling for the stability of each variable across time.

Cross-lagged and reciprocal effects models tested links between child symptomatology (depression and DBD) and maternal depression symptom levels (all continuous measures). Subgroup comparisons using stacked modeling procedures ( Bollen, 1989 ) were used to assess whether the magnitude of parameter estimates differed in strength for boys and girls. Post hoc Wald tests were used to test the assumption of equality between the targeted paths (effect of child symptomatology on maternal depression symptoms) across gender. A significant Wald test statistic provides an estimate of the statistical significance of any specific parameter comparisons (i.e., whether there are any differences in effect by child gender).

Past maternal depression severity was included as a covariate in all analyses. Note that in structural equation modelling (SEM) analyses, covarying for differences in the constructs at the first measurement occasion accounts for the influence of the covariate at the later time point via the various indirect pathways ( Little, Preacher, Selig, & Card, 2007 ). To test whether results were influenced by covarying patterns of social disadvantage, sensitivity analyses further adjusted for household income.

Structural equation model analyses were undertaken in Mplus 7.11 ( Muthén & Muthén, 1998 ) using robust maximum likelihood estimation procedures.

Girls

Figure 1a shows results for the cross-lagged model linking girls’ depression symptoms and maternal depression symptom levels after adjusting for maternal past depression severity and child age. The stability coefficients across time for maternal depression and girls’ depression symptoms were moderate (β = .41, 95% CI [.28, .54], p < .001; β = .30, 95% CI [.14, .46], p < .001, respectively) and did not significantly differ from each other. There was evidence of a moderate effect of girls’ depression symptoms at baseline on maternal depression symptoms at follow-up (β = .16, 95% CI [.03, .40], p = .039), but not between maternal depression levels (baseline) and girls’ depression symptoms at follow-up (β = .07, 95% CI [−.09, .23], p = .376), suggesting that girls’ depression symptoms were associated with an increase in maternal depression symptom levels across time, but not vice versa. Tests of reciprocal effects models ( Figure 1b ) demonstrated no evidence of influences of girls’ depression symptoms on maternal depression symptom levels (β = .12, 95% CI [−.25, .48], p = .602), or vice versa (β = −.06, 95% CI [−.45, .32], p = .748).

The Wald test examined differences between male and female offspring by comparing specific parameters between boys and girls. The path from child depression symptoms (baseline) to maternal depression symptoms (follow-up) differed for boys and girls (Wald (1) = 3.74, p = .042).

Analyses were also conducted using complete case analysis and are presented in supplemental Figure 1 (panels a and b). The pattern of results was similar; however, the effect of girls’ depression symptoms on maternal depression was slightly weaker (β = .14, 95% CI [−.03, .30], p = .108).

Offspring Disruptive Behaviors

Figure 1c shows cross-lagged models for offspring DBD symptoms and maternal depression symptoms. No longitudinal cross-lagged effects were observed between boys’ DBD symptoms and maternal depression symptoms. There was weak evidence of an effect of girls’ DBD symptoms on maternal depression symptoms, but this association fell below conventional levels of significance (β = .12, 95% CI [−.03, .26], p = .096). Tests of reciprocal effects models ( Figure 1d ) demonstrated no evidence of influences of boys’ or girls’ DBD symptoms on maternal depression symptom levels, or vice versa.

Analyses were also conducted using complete case analysis and are presented in Figure 1 (panel c and d). The pattern of results was the same.

Sensitivity analyses also examined household income as an additional potential confounder. Approximate gross family income was assessed by parent report questionnaire at baseline. Household income was coded on a scale of 1 (<£10,000) to 7 (>£60,000), with lower scores indicating lower household income (see Table 3 ). Income was associated with maternal depression symptoms (boy model: β = −.10, 95% CI [−.18, −.02], p = .015; girl model: β = −.05, 95% CI [−.12, .01], p = .090) but not child symptoms of depression (boys: β = −.03, 95% CI [−.09, .02], p = .291; girls: β = −.04, 95% CI [−.09, .01], p = .141) or DBD (boys: β = −.04, 95% CI [−.12, .02], p = .224; girls: β = −.02, 95% CI [−.07, .03], p = .497). Sensitivity analyses additionally covaried for household income in cross-lagged models. The pattern of results remained unchanged with girls’ symptoms of depression at baseline remaining associated with maternal depression symptoms at follow-up (β = .17, 95% CI [.01, .35], p = .050) and weak evidence of effects of girls’ DBD symptoms at baseline on maternal depression symptoms at follow-up (β = .14, 95% CI [−.02, .30], p = .083).

Discussion

A number of important studies have highlighted risk to offspring from having a parent with depression ( Beardslee, Versage, & Gladstone, 1998 ; Brennan et al., 2000 ; Weissman, Warner, Wickramaratne, Moreau, & Olfson, 1997 ). The current study took a different approach by examining the extent to which child psychopathology impacts on maternal depressive symptom course and illness recurrence. Findings indicate that in mothers with a previous history of depressive disorder, offspring depression symptoms during adolescence predict an exacerbation of depressive illness in mothers. Mothers with a past history of recurrent depression experienced a 9% increase in risk for a further episode of depression at follow-up for each additional offspring symptom of depression reported at study baseline after accounting for background risk (household income and past maternal depression severity). Structural equation models provided further evidence of risk effects of offspring depression on the course of maternal depression, particularly so in the case of mother–daughter dyads. Furthermore, this association was not due to shared rater variance as children reported on their own symptoms. Findings for effects of DBD on maternal depression symptom course were inconclusive, with only weak evidence of an association between girls’ DBD symptoms (at baseline) and maternal depression symptoms at follow-up.

It is well-established that maternal depression is associated with increased risk for offspring psychopathology when compared with children of nondepressed mothers ( Goodman & Gotlib, 1999 ), and the rate of offspring psychiatric disorder in this sample was high (24%) when compared with general population rates (∼11%; Green, McGinnity, Meltzer, Ford, & Goodman, 2004 ). In keeping with previous findings from this sample ( Mars et al., 2012 ), maternal past depression severity was also associated with variation in offspring depression symptoms within this high-risk group.

It is important to examine whether child psychiatric symptoms have an adverse effect on the course of maternal depression to better understand risk factors that underlie the maintenance or recurrence of adult depression. We extended previous research by examining this in a sample of mothers with a history of recurrent episodes of major depression. Findings showed that offspring symptoms of depression (but not DBD) predicted a future episode recurrence in mothers, even when accounting for prior maternal depression severity. Cross-lagged and reciprocal effects models provided further and more stringent tests of possible transactional effects between child and maternal psychopathology over time by simultaneously taking into account associations between variables at baseline and follow up, and considering the stabilities of both maternal depression and child symptomatology. Therefore, although the main focus of the study was to examine child effects on maternal mental health, the modeling allows for baseline maternal depression symptom levels to predict change in child outcomes. Furthermore, by examining the association between child and maternal symptoms (using a cross-lagged approach), we also examined whether child gender moderated observed associations. Girls’ symptoms of depression predicted an increase in maternal depression levels over time using cross-lagged models, and effects for girls were greater than for boys. Findings were not explained by cross-sectional correlation of maternal and child symptoms at baseline, by previous maternal depression severity, or by variation in household income.

Findings of gender differences are in accord with a growing body of literature suggesting that maternal depression may be more strongly associated with girls’, rather than with boys’, depression ( Davies & Windle, 1997 ; Fergusson, Horwood, & Lynskey, 1995 ). Previous studies have suggested that girls’ symptoms of depression and DBD predict maternal depression symptoms in the general population ( Guo & Slesnick, 2011 ; Jaffee & Poulton, 2006 ), although findings of so-called child effects on parent have been mixed. We focused on adult patients who had been treated in primary care. For this group, our findings suggest that the presence and impact of daughters’ depression symptoms on mothers is important to consider when examining risk factors that contribute to the course of depression in women during offspring adolescence.

Although the focus of this study was to test child effects on parents, it is important to note that, in contrast to some previous research ( Allen et al., 2010 ; Hughes & Gullone, 2010 ; Kouros & Garber, 2010 ; Nicholson et al., 2011 ), we did not observe bidirectional longitudinal effects between maternal depression levels and later offspring DBD or depression. This is not to suggest that maternal depression as a risk for offspring psychopathology is not important; the study was not designed to test this, as control participants were not included. This is because maternal depression is a well-established risk factor for offspring psychopathology ( Brennan, Hammen, Katz, & Le Brocque, 2002 ; Lieb, Isensee, Höfler, Pfister, & Wittchen, 2002 ; Wickramaratne & Weissman, 1998 ). One possible explanation for why we did not observe effects of maternal depression symptom levels on change in offspring psychopathology is that the current study utilized a high-risk sample of parents, all of whom had a long history of recurrent depressive illness, and other aspects of maternal depression may be more important in terms of predicting variation in risk for the development of psychopathology within such high-risk samples. For example, severity, chronicity and recurrence of maternal depression episodes have been found to impact on risk for offspring psychopathology in this and other high-risk samples ( Hammen & Brennan, 2003 ; Mars et al., 2012 ). Because of all offspring being exposed to some level of maternal depression, it may be less likely

Clinical Implications

The findings of the current study suggest that for clinicians who deal with adult depression, asking about offspring psychopathology is important and relevant to the assessment and ongoing management of depression, particularly given the relatively high risk of depression in women of child-rearing age. In addition, improved communication and coordination between different services dealing with adult depression and children’s mental health may assist with more effective monitoring and management of mental health problems in affected families. In particular, there are a range of nonhealth care agencies providing care to children and adolescents with mental health problems (such as social services, youth offending services, and education), and often little information from these agencies is available to clinicians dealing with adult mental health problems. Effective prevention and treatment of offspring psychopathology is an important priority in its own right, and findings from this study highlight that there may be additional benefits for the management of adult depression.