- Email: [email protected]
Perceived parenting pattern and response to antidepressants in patients with major depression
Journal of Affective Disorders 52 (1999) 59–66
Research report
Perceived parenting pattern and response to antidepressants in patients with major depression a, b a a a Kaoru Sakado *, Tetsuya Sato , Toru Uehara , Miwako Sakado , Toshiyuki Someya
b
a Department of Psychiatry, Niigata University School of Medicine, Niigata, Japan Department of Psychiatry, Fujita Health University School of Medicine, Toyoake, Japan
Received 27 October 1997; received in revised form 15 April 1998; accepted 15 April 1998
Abstract Background: No systematic study has been conducted to explore the relationship of dysfunctional parenting early in life, as measured by the Parental Bonding Instrument (PBI), to outcomes of depression, although a number of studies have related parenting behaviors to the development of depression in adulthood. Methods: The relationship between PBI scores and 4-month outcomes after treatment with antidepressants was explored in 60 outpatients with major depression, controlling for potentially confounding factors. Results: A multiple logistic regression analysis suggested that low levels of paternal care, unmarried condition, non-melancholic features, and a high isolation tendency were all factors that contributed to poor outcomes for depression. The predictive power of low paternal care was not influenced by levels of depression or neuroticism. Limitation: This study did not attempt to explore whether the effects of parenting of father and mother on outcomes for depression may differ between male and female subjects. Conclusion: The results suggest that low levels of paternal care may be an independent predictor of a poor response to treatment with adequate antidepressants. 1999 Elsevier Science B.V. All rights reserved. Keywords: Parental bonding instrument; Major depression; Treatment outcome; Personality
1. Introduction Theorists and clinicians have long assumed that exposure in childhood to dysfunctional parenting is a
*Corresponding author. Tel.: 1 81-25-2272213; fax: 1 81-252270777. E-mail address: [email protected] (K. Sakado)
risk factor for depression in adulthood (see Becker, 1974; Parker, 1983a). Subsequent research in this area has led to a proposal of two principal factors in perceived child–parent relationships which may be relevant to depression in adulthood: lack of warmth or care and negative control (overprotection) (Parker et al., 1979; Raskin et al., 1971). Since the first report by Parker (1979), a large number of casecontrol and community studies (Kerver et al., 1992;
0165-0327 / 99 / $ – see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S0165-0327( 98 )00062-7
60
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
Mackinnon et al., 1993; Parker, 1983b; Parker et al., 1995; Parker, 1992; Plantes et al., 1988; Rodgers, 1996; Sato et al., 1997) have empirically confirmed that these two factors, particularly when measured by the Parental Bonding Instrument (PBI), can be associated with the development of depression or a depressive state in adulthood (particularly nonmelancholic or reactive depression). It has been suggested that the PBI can be correlated with not only onset, but also the course or outcome of depression. Gotlib et al. (1988) followed depressed postpartum women for an average of 30 months, and found that subjects who had received a lower level of parental care were still symptomatic at the last follow-up. They therefore have concluded that the level of perceived care is predictive of levels of depression later in life. However, there are some limitations in these findings. First, because the subjects in their study were comprised of depressed postpartum women as judged by the current level of the Beck Depression Inventory (Beck et al., 1961), it is questionable whether similar findings will be found in subjects diagnosed as having a depressive disorder. Second, it is unclear whether the subjects received treatment during the follow-up period, and if they had, it is not known how they were treated. In addition, the study did not examine the potentially confounding effects of other psychosocial factors (such as personality pathology), which are wellknown to influence the outcome of depression. Unfortunately, no systematic study has investigated the relationship between perceived parental characteristics and the outcome of depression, if the limitations of the Gotlib et al. (1988) study are considered carefully. The present study investigates the influence of parenting behaviors and attitudes, as measured by the PBI, on 4-month outcomes for depression. To minimize the limitations noted in the study described above, this study was designed as follows: we chose outpatients diagnosed as having major depression according to the Diagnostic and Statistical Manual of Mental Disorders, 3rd edition, revised (DSM-III-R). The subjects were followed closely while receiving adequate doses of antidepressants. Logistic regression analysis was used to separate out the effects of personality pathologies as well as demographic or
clinical characteristics on the outcomes for depression.
2. Subjects and methods
2.1. Materials and procedure The sample group consisted of new patients who were consecutively admitted to the outpatient clinic of the Psychiatric Department at Niigata City General Hospital from April 1994 to January 1995. After informed consent was obtained, the Structured Clinical Interview for DSM-III-R (SCID; Spitzer et al., 1990) was administered to depressed patients within 1 week of their first clinical examination. The patients diagnosed with major depression by SCID were included in the study. Patients with psychotic features, a lifetime history of hypomanic or manic episodes, panic or generalized anxiety disorders, alcohol or substance abuse, or organic brain syndrome were excluded. Patients younger than age 21 or older than age 65 were also excluded. A total of 69 patients met the criteria for initial inclusion. The patients were prescribed adequate levels of tricyclic antidepressants for at least 20 weeks. Either imipramine, clomipramine, amitriptyline, or maprotiline was used for each patient, with small doses of benzodiazepines and / or levomepromazine. The efficacy of maprotiline is believed to be equal to that of tricyclic antidepressants (Logue et al., 1979). If there was an inadequate response to the antidepressants, dosages were increased every week during the initial 8 weeks up to the maximum allowed by the Japan Pharmaceutical Information Center (1995) (200mg / day for imipramine; 225mg / day for clomipramine; 150mg / day for amitriptyline; 200mg / day for maprotiline). If a patient’s response to the medication was still inadequate, the maximum dosage was used for the remainder of the study (at least 12 weeks). During the follow-up period, six patients dropped out. Two patients who required inpatient treatment were referred to other hospitals, one patient underwent a manic episode, and three patients discontinued treatment. In addition, three patients were excluded because they could rate only one parent on
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
the PBI scale, due to the death of the other parent. The remaining 60 patients were included in the study. Of the 60 patients, 30 (50%) were male and 30 (50%) were female. The mean age was 43.0 (s.d. 5 12.0; range 22–64) years. Thirteen patients (22%) were unmarried and 47 patients (78%) were married. The mean of education was 12.0 (s.d. 5 2.4; range 8–16) years. Ten patients (17%) had no regular occupation. The mean score on the 17-item Hamilton Rating Scale for Depression (HAM-D; Hamilton, 1960) at baseline was 19.2 (s.d. 5 5.8; range 12–38). The mean age of onset for the first depressive episode was 41.0 (s.d. 5 12.6; range 16– 63) years. The number of patients with recurrent major depression was 18 (30%). The depressive state for 35 (58%) of the patients met the criteria for the melancholic subtype, according to the DSM-III-R. Of the 60 patients, 11 (18%) had a family history of affective disorders. There was no significant difference between the 60 patients included in this study and the 9 patients excluded during the follow-up period in terms of the sex ratio (% of female: 50 v. 44%), mean age (43.0 v. 43.6 years), marital status (% of unmarried: 22 v. 22%), mean education (12.0 v. 13.3 years), occupation (% of no regular occupation: 17 v. 11%), HAM-D score at baseline (19.2 v. 17.6), mean onset age of first depressive episode (41.0 v. 43.0), number of patients with recurrent major depression (30 v. 22%), melancholic features (% of melancholic subject: 58 v. 67%), or family history of affective disorders (% of affective disorders: 18 v. 22%). The severity of depression of the patients was evaluated three different times using the HAM-D: at the baseline, after 8 weeks, and after 16 weeks. At the second assessment of depressive symptoms, the subjects also completed the PBI and MPT (von Zerssen et al., 1988), which is a short-item selfrating questionnaire assessing six personality dimensions (see below). With respect to the MPT, subjects were asked to respond to the questionnaire based on their assessment of themselves in a physically and mentally healthy state. The 4-month outcome (remitted or non-remitted) was determined for each patient according to the following criteria of remission: (1) less than eight points on the HAM-D, as proposed by Frank et al.
61
(1991); (2) a full recovery in social function at the 16-week follow-up; and (3) no sign of recurrence of depression during the 4 weeks after the assessment according the two criteria listed above. Outcomes were assessed by one of us (K.S. or T.U.), both of us being blind to both the clinical data at baseline and scores on the PBI and MPT.
2.2. Measures The PBI is a self-report questionnaire developed by Parker et al. (1979). Respondents answer 25 questions about each parent, based on their memories of parents during the first 16 years of their childhood. Using this questionnaire, two child-rearing styles (care and protection) can be assessed for each parent. These parenting styles are derived by factor analysis. In addition, it has been suggested that the PBI can measure with validity actual, and not merely perceived, parental behavior (Parker, 1992). Kitamura and Suzuki (1993) has developed the Japanese version of the PBI using the back-translation method. Their validation study has found that the factor structure of the Japanese version is similar to that of the original questionnaire. In the present study, we used the Japanese version of the questionnaire to assess parenting styles. To assess the personality of the subjects, the MPT (Munich Personality Test; von Zerssen et al., 1988) was used. The MPT is a self-rating questionnaire of 51 items designed to assess six personality dimensions: Neuroticism, Extraversion, Frustration Tolerance, Rigidity, Isolation Tendency, and Esoteric Tendencies. Neuroticism and Extraversion in the MPT are nearly identical to the corresponding concepts advocated by Eysenck (1959). Frustration Tolerance primarily measures traits like emotional stability and insensitivity. Rigidity represents a kind of obsessionality (love of orderliness, dependence on norms and perfectionism, and so on). Both the Isolation Tendency and Esoteric Tendencies describe personality pathologies similar to schizoid and schizotypal personality disorders defined in the DSM-III-R. We used the Japanese version of the MPT, which was translated in keeping with the back-translation method, and has shown a high level
62
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
of consistency with the original test (Sakado et al., 1996).
2.3. Analysis To determine whether assessments of recollections of parenting behaviors and personality were independent of the level of depression, correlation of the HAM-D score at the 8-week follow-up, when the PBI and MPT were administered, to the PBI and MPT dimensional scores was calculated prior to other analyses. The sample was divided into remitted and nonremitted patients on the basis of the 4-month outcome assessment. The demographic and clinical data were univariately compared between the two groups. Similarly, mean scores on each dimension of both the PBI and the MPT were compared between the two groups. In addition, to examine whether or how these variables contributed to the outcome, a multiple logistic regression procedure was performed, using any variables which reached a significant level in the above analyses. To obtain the most appropriate model, Akaike’s information criterion (AIC; Akaike, 1987) was calculated, and a combination of the variables producing the minimum AIC was selected. For the analyses, a computer package, ‘‘SPSS for Windows 6.1J’’ was used. The significant level was set at 0.05 throughout this study.
3. Results
3.1. Correlation of MPT and PBI dimensional scores to depressive symptoms The mean HAM-D score at the 8-week follow-up, when the PBI and MPT were administered, was 3.6 (s.d. 5 3.4), suggesting that assessments of recollection of parenting behaviors and personality were independent of depressive level. This was confirmed by calculating the correlation coefficients of the HAM-D score at the 8-week follow-up to the PBI and MPT dimensional scores. Coefficients were: 2 0.16 (P 5 0.23) for paternal care, 0.18 (P 5 0.17) for paternal protection, 0.14 (P 5 0.30) for maternal
care, 0.02 (P 5 0.89) for maternal protection, 0.19 (P 5 0.18) for Extraversion, 0.16 (P 5 0.22) for Neuroticism, 0.03 (P 5 0.80) for Frustration Tolerance, 2 0.01 (P 5 0.91) for Rigidity, 0.03 (P 5 0.84) for Isolation Tendency, and 0.19 (P 5 0.18) for Esoteric Tendencies. No coefficients were significant. There was a high and significant correlation between paternal and maternal care scores (r 5 0.48, P 5 0.000) and between paternal protection and maternal protection scores (r 5 0.65, P 5 0.000) in the sample.
3.2. Univariate analysis Thirty six patients (60%) were judged to have remitted and 24 (40%) had not remitted at the 4-month follow-up. The demographic and clinical characteristics are shown in Table 1, together with the comparison between the remitted and non-remitted groups. In this comparison, some significant differences were found between the two groups. The mean age in the remitted group was significantly higher than in the non-remitted group (t 5 2 2.7, d.f. 5 58, P 5 0.009). The ratio of patients unmarried in the remitted group was significantly lower than in the non-remitted group ( x 2 5 5.9, d.f. 5 1, P 5 0.02). The ratio of patients with melancholic features in the remitted group was significantly higher than in the non-remitted group ( x 2 5 7.1, P 5 0.008). There were no other significant differences in demographic and clinical variables between the two groups. Table 2 shows the scores on each PBI or MPT dimension for both the remitted and non-remitted groups. A comparison between the two groups is also given. On the PBI, the mean score for paternal care in the remitted group was significantly higher than in the non-remitted group (t 5 2 2.2, d.f. 5 58, P 5 0.03). On the MPT, the remitted group showed significantly lower mean scores for Extraversion (t 5 2 2.3, d.f. 5 58, P 5 0.03), Neuroticism (t 5 2 2.8, d.f. 5 58, P 5 0.008), Isolation Tendency (t 5 2 2.8, d.f. 5 58, P 5 0.008), and Esoteric Tendencies (t 5 2 2.5, d.f. 5 34.9, P 5 0.02) than did the non-remitted group.
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
63
Table 1 Demographic and clinical characteristics of the subjects Variables
Sex ratio, female, n (%) Age, mean (s.d.), years Education, mean (s.d.), years Marital status, n (%) Married Unmarried No regular occupation, n (%) HAM-D, mean (s.d.) Onset, mean (s.d.), years Number of patients with recurrent major depression, n (%) Melancholic features according to DSM-III-R, n (%) Family history of major affective disorders, yes, n (%)
All subjects
Remitted
Non-remitted
Comparison between groups
(n 5 60)
(n 5 36)
(n 5 24)
t or x 2
d.f.
P
95% CI
x 5 0.0 t 5 2.7 t 5 2 1.3 x 2 5 5.9
1 58 58 1
NS (1-tailed) P 5 0.009 (2-tailed) NS (2-tailed) P 5 0.02 (1-tailed)
2 0.3 to 0.3 2.2 to 14.2 2 2.1 to 0.4 0.04 to 0.5
2 0.1 to 0.3 2 2.8 to 3.6 2 0.3 to 12.7 2 0.2 to 1.72
2
30 (50) 43.7 (13.8) 12.0 (2.4)
18 (50) 46.3 (10.8) 11.7 (2.4)
12 (50) 38.1 (12.3) 12.5 (2.3)
47 (78) 13 (22) 10 (17) 19.2 (5.8) 41.0 (12.6) 18 (30)
32 (89) 4 (11) 4 (11) 19.4 (6.1) 43.6 (12.7) 11 (31)
15 (63) 9 (38) 6 (25) 19.0 (6.0) 37.4 (29) 7 (29)
Fisher’s exact test t 5 0.2 t 5 1.9 x 2 5 0.01
58 58
NS NS NS NS
35 (30)
26 (31)
9 (38)
x 2 5 7.1
1
P 5 0.008 (1-tailed)
0.1 to 0.6
11 (18)
9 (25)
2 (8)
x 2 5 2.7
1
NS (1-tailed)
2 0.01 to 0.3
(2-tailed) (2-tailed) (2-tailed) (1-tailed)
Table 2 Comparison of PBI and MPT between remitted and not remitted groups
PBI Paternal care Paternal protection Maternal care Maternal protection MPT Extraversion Neuroticism Frustration tolerance Rigidity Isolation tendency Esoteric tendencies
(A)
(B)
Comparison between groups
Remitted (n 5 36) mean (s.d.)
Non-remitted (n 5 24) mean (s.d.)
t
d.f.
P (2-tailed)
95% CI
23.7 11.2 23.7 11.2
(5.9) (6.4) (6.4) (6.8)
19.8 12.3 24.3 11.1
(7.8) (6.1) (5.8) (8.0)
2.2 2 0.6 2 0.4 0.0
58 58 58 58
0.03 NS NS NS
0.3 to 7.4 2 4.3 to 2.3 2 3.9 to 2.6 2 3.8 to 3.9
A.B
8.0 10.1 5.0 11.1 3.6 1.7
(4.5) (4.7) (3.3) (4.3) (2.0) (1.6)
10.7 14.1 4.7 12.0 5.2 3.1
(4.4) (6.5) (3.7) (5.4) (2.7) (2.5)
2 2.3 2 2.8 0.3 2 0.7 2 2.8 2 2.5
58 58 58 58 58 34.9
0.03 0.008 NS NS 0.008 0.02
2 5.0 to 2 6.9 to 2 1.5 to 2 3.4 to 2 2.9 to 2 2.6 to
A,B A,B
2 0.3 2 1.1 2.1 1.6 2 0.4 2 0.3
A,B A,B
PBI, Parental Bonding Instrument; MPT, Munich Personality Test; CI, confidence interval.
3.3. Multiple logistic regression analysis The univariate analyses revealed that three of the demographic and clinical variables (age, marital status, and melancholic features), one of the PBI dimensional scores (paternal care), and four of the MPT dimensional scores (Extraversion, Neuroticism, Isolation Tendency, and Esoteric Tendencies) may have been predictive of the 4-month outcome for the
subjects. These eight variables were therefore considered as possible candidates for a logistic model that could best fit the data. Although a stepwise procedure is often used to identify subsets of independent variables that are good predictors of the dependent variable, there is a possibility that this procedure may not yield a ‘‘best’’ variable set. Therefore, all combinations of those eight variables were explored by performing 255 (2 8 2 1) logistic
64
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
regression analyses with the 4-month outcome as the dependent variable, and comparisons of model fits were made by calculating AIC for each logistic model. AIC is known to reach a minimum when the proposed model best fits the data (Akaike, 1987). The AIC calculations in this study demonstrate that a combination of four variables (marital status, melancholic features, paternal care of the PBI, and Isolation Tendency of the MPT) gave the minimum AIC value, indicating that the logistic model including these four variables was the most appropriate in predicting the 4-month outcome. The equation for this logistic model is shown in Table 3. The AIC for the model was 65.5, which was significantly lower than the null model including only a constant (AIC 5 80.8; x 2 5 25.2, d.f. 5 4, P 5 0.0000). In the equation, all four variables are significant in predicting the 4-month outcome. The odds ratios for having a poorer treatment outcome were: 6.0 for marital status (unmarried subjects against married subjects), 6.9 for melancholic features (subjects with non-melancholic features against those with melancholic features), 1.9 for paternal care (subjects reporting a five-point lower score), and 2.5 for Isolation Tendency (subjects reporting a three-point higher score), and all were significant. The odds ratios were judged to be significant if their 95% confidence intervals (CIs) did not include 1.0. It has been suggested that PBI scores may be influenced by response biases caused by the level of depression or trait characteristics such as neuroticism (Parker, 1983b). Correlation analyses in this study, however, show that the PBI scores are independent of the level of depression. We also attempted to
determine whether neuroticism has an influence on the relationship between low paternal care and poor treatment outcome indicated by the above logistic regression analysis. The Neuroticism score from the MPT was entered into the best-fitting model described above, and the predictive values of the variables included in the equation were observed. Using the Neuroticism score, the model gave an AIC value of 67.4, indicating that this model is a poorer fit for the data than that described in Table 3. The odds ratios for having a poorer treatment outcome were: 5.0 (NS) for marital status (unmarried subjects against married subjects), 6.7 (P , 0.05, 95% CI 5 1.6–27.7) for melancholic features (subjects with non-melancholic features against those with melancholic features), 1.9 (P , 0.05, 95% CI 5 1.7–2.2) for paternal care (subjects reporting a five-point lower score), 2.3 (NS) for Isolation Tendency (subjects reporting a three-point higher score), and 1.1 (NS) for Neuroticism (subjects reporting a threepoint higher score). Probably because there is a high correlation between Neuroticism and marital status (r 5 0.47, P 5 0.000 when treating marital status as a dummy variable), and between Neuroticism and Isolation Tendency (r 5 0.41, P 5 0.001), these two variables lose much of their predictive power in the equation. However, non-melancholic features and paternal care remained significant.
4. Discussion The present study attempted to explore whether parenting behaviors, as measured by the PBI, in-
Table 3 Results of multiple logistic regression analysis Independent variables Dependent variable 0 5 remitted 1 5 non-remitted AIC 5 65.5 Model R 2 5 0.46 Unmarried Non-melancholic features Paternal care Isolation Tendency a
b
Odds ratio
95% CI
Wald statistics
1.8 1.9 2 0.1 0.3
6.0 a 6.9 a 1.9 (for a 5-point score lower)a (2.5 (for a 3-point score higher)a
1.2 to 30.3 3.4 to 13.8 1.7 to 2.2 1.9 to 3.4
4.7 (P 5 0.03) 7.1 (P 5 0.008) 5.2 (0.02) 3.9 (0.048)
Significant odds ratio. Subjects reporting a 5-point lower score for paternal care are 1.9 times more likely to have a poorer outcome. AIC, Akaike’s Information Criterion; CI, confidence interval. b
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
fluence short-term outcomes for depressive outpatients treated with adequate antidepressants. Patients with poorer treatment outcomes reported a significantly lower score for paternal care than did those who responded well to the antidepressants, and this difference remained significant even when the possible effects of demographic, clinical, and personality variables were separated out. The multiple logistic regression analyses conducted in this study demonstrate that subjects reporting a five-point lower score for paternal care were 1.9 times more likely to have a poorer response to treatment with adequate antidepressants. The correlation analyses conducted for this study indicate that the PBI scores were independent of the level of depression. In the logistic model which includes Neuroticism, paternal care remains a significant predictor of response to antidepressants. These results indicate that the association between low paternal care and a poor response to antidepressants are not due to response biases caused by the level of depression or trait characteristics such as neuroticism, suggesting that there may be a causal linkage. One might expect that a low score for paternal care would be associated with non-melancholic features of depression, as previous studies on the PBI in depressive patients have suggested (Parker, 1992; Parker and Hadzi-Pavlovic, 1993). This association, it would seem, could account for the significance of low paternal care in predicting a poor response to antidepressants. However, the most appropriate logistic model including both paternal care and nonmelancholic features shows that paternal care is significant in predicting the poor response even after the effects of the non-melancholic features are separated out. The results of this study suggest that low paternal care is a significant risk factor for a poor response to treatment with adequate antidepressants. This risk appears to be independent of nonmelancholic features, personality assessments, and other demographic variables. Although Gotlib et al. (1988) were the first to find that low parental care may contribute to poor outcomes for depression, some methodological limitations in their study, as mentioned above, make it difficult to interpret their findings. The present study, which used a more clearly defined sample and was better controlled for treatment conditions and other
65
potentially confounding factors, provides more direct evidence for an independent link between low parental care and outcomes for depression. The results of this study may enhance researchers’ interest in the association between parenting behaviors and depression. Most previous studies in this area have explored parenting behavior as a cause of depressive disorders in adulthood. Evidence derived from these studies suggests that low parental care rather than high parental protection may be a primary risk factor for depressive disorders (Mackinnon et al., 1993; Parker, 1983a). It is still controversial, however, whether paternal or maternal care is more influential in the development of depressive disorders (Rodgers, 1996). This study indicates that only low paternal care (but not maternal care) is significant in predicting poor treatment outcomes in patients with major depressive disorder, although there does appear to be a high correlation between paternal and maternal care scores. These results suggest that low paternal care may be relevant not only to the pathogenesis of depressive disorders but also to poor treatment outcomes for depression, and that the influence on these outcomes may differ when considering levels of paternal and maternal care. Similar studies are needed to replicate and confirm the findings of this study. Although this study found no sex difference in the outcomes between the remitted and non-remitted groups, there is a possibility that the effects of parenting of father and mother on outcomes for depression may be different between male and female subjects. Unfortunately, we could not analyze the data after separating the subjects into men and women, because our sample size was somewhat small. Naturally, a further study is needed to explore whether the relationship of the PBI to outcomes for depression differs between male and female subjects. Based on the results of this study, it is unclear how low paternal care is causally linked to a poor response to antidepressants. Parker et al. (1992) have reviewed previous studies investigating the continuity between parenting behavior, as measured by the PBI, and social bonds in adulthood. They have pointed out that there is some evidence for a link between low parental care and deficiencies in diffuse social bonds. In addition, some studies have suggested that depressive patients with deficiencies in
66
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
social bonds have a poorer response to antidepressants (Lam et al., 1994). One can speculate that deficiencies in social bonds may be responsible for the causal linkage between low parental care and a poor response to antidepressants. On the other hand, considering that Kendler (1996) found evidence of the effect of genetic factors on parental care, a poor response to antidepressants in subjects who received low parental care may be influenced by genetic factors.
References Akaike, H., 1987. Factor analysis and AIC. Psychometrika 52, 317–332. Beck, A.T., Ward, C.H., Mendelson, M., et al., 1961. An inventory for measuring depression. Arch. Gen. Psychiatry 4, 561–571. Becker, J., 1974. Depression. Theory and Research. Winston, Washington, DC. Eysenck, H.J., 1959. Manual of the Maudsley Personality Inventory. University of London Press, London. Frank, E., Prien, R.F., Jarrett, R.B., et al., 1991. Conceptualization and rationale for consensus definitions of terms in major depressive disorder. Remission, recovery, relapse, and recurrence. Arch. Gen. Psychiatry 48, 851–855. Gotlib, I.H., Mount, J.H., Cordy, N.I., et al., 1988. Depression and perceptions of early parenting: a longitudinal investigation. Br. J. Psychiatry 152, 23–27. Hamilton, M., 1960. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23, 56–62. Japan Pharmaceutical Information Center, 1995. Drug in Japan: Ethical Drugs. Yakugyojiho-Sya, Tokyo. Kendler, K.S., 1996. Parenting: a genetic-epidemiological perspective. Am. J. Psychiatry 153, 11–20. Kerver, M.J., van Son, M.J.M., de Groot, P.A., 1992. Predicting symptoms of depression from reports of early parenting: a one-year prospective study in a community sample. Acta Psychiatr. Scand. 86, 267–272. Kitamura, T., Suzuki, T., 1993. A validation study of the Parenting Bonding Instrument in a Japanese population. Jap. J. Psychiatr. Neurol. 47, 29–36. Lam, D.H., Green, B., Power, M.J., et al., 1994. The impact of social cognitive variables on the initial level of depression and recovery. J. Affect. Disord. 32, 75–83. Logue, J.N., Sachais, B.A., Feigher, J.P., 1979. Comparisons of maprotiline with imiparamine in severe depression: a multicenter controlled trial. J. Clin. Pharmacology. 19, 64–74.
Mackinnon, A.J., Henderson, A.S., Andrew, G., 1993. Parental ‘‘affectionless control’’ as an antecedent to adult depression: a risk factor refined. Psychol. Med. 23, 135–141. Parker, G., Tupling, H., Brown, L.B., 1979. A Parental Bonding Instrument. Br. J. Med. Psychol. 59, 1–10. Parker, G., 1979. Parental characteristics in relation to depressive disorders. Br. J. Psychiatry 134, 138–147. Parker, G., 1983a. Parental Overprotection: A Risk Factor in Psychosocial Development. Grune and Stratton, New York, London, Paris, San Diego, San Francisco, Sao Paulo, Sydney, Tokyo and Toronto. Parker, G., 1983b. Parental ‘‘affectionless control’’ as an antecedent to adult depression. Arch. Gen. Psychiatry 40, 956–960. Parker, G., 1992. Early environment. In: Paykel, E.S. (Ed.), Handbook of Affective Disorders, 2nd ed. Churchill Livingstone, Edinburgh, London, Madrid, Melbourne and Tokyo. Parker, G., Barrett, E., Hickie, I.B., 1992. From nurture to network: examining links between perceptions of parenting received in childhood and social bonds in adulthood. Am. J. Psychiatry 149, 877–885. Parker, G., Hadzi-Pavlovic, D., 1993. Parental representation of melancholic and non-melancholic depression: examining for specificity to depressive type and for evidence of additive effects. Psychol. Med. 22, 657–665. Parker, G., Hadzi-Pavlovic, D., Greenwald, S., et al., 1995. Low parental care as a risk factor to lifetime depression in a community sample. J. Affect. Disord. 33, 173–180. Plantes, M.M., Prusoff, B.A., Brennan, J., et al., 1988. Parental representations of depressed outpatients from a U.S.A. sample. J. Affect. Disord. 15, 149–155. Raskin, A., Boothe, H.H., Reatig, N.A., et al., 1971. Factor analyses of normal and depression patients’ memories of parental behavior. Psychiatr. Rep. 29, 871–879. Rodgers, B., 1996. Reported parental behavior and adult affective symptoms. 1. Associations and moderating factors. Psychol. Med. 26, 51–61. Sakado, K., Sato, T., Uehara, T., et al., 1996. The Japanese version of the Munich Personality Test (MPT): its reliability and validity. Arch. Psychiatr. Diag. Clin. Eval. 7, 123–132. (in Japanese). Sato, T., Sakado, K., Uehara, T., et al., 1997. Perceived parental style in a Japanese sample of depressive disorders. A replication outside Western culture. Br. J. Psychiatry 170, 173–175. Spitzer, R., Williams, J., Gibbon, M. et al., 1990. Structured Clinical Interview for DSM-III-R (SCID). User’s Guide. APA, Washington, DC. von Zerssen, D., Pfister, H., Kleller, D.M., 1988. The Munich Personality Test (MPT). A short questionnaire for self-rating and relatives’ rating of personality traits: formal properties and clinical potential. Eur. Arch. Psychiatr. Neurol. Sci. 238, 73– 93.
Research report
Perceived parenting pattern and response to antidepressants in patients with major depression a, b a a a Kaoru Sakado *, Tetsuya Sato , Toru Uehara , Miwako Sakado , Toshiyuki Someya
b
a Department of Psychiatry, Niigata University School of Medicine, Niigata, Japan Department of Psychiatry, Fujita Health University School of Medicine, Toyoake, Japan
Received 27 October 1997; received in revised form 15 April 1998; accepted 15 April 1998
Abstract Background: No systematic study has been conducted to explore the relationship of dysfunctional parenting early in life, as measured by the Parental Bonding Instrument (PBI), to outcomes of depression, although a number of studies have related parenting behaviors to the development of depression in adulthood. Methods: The relationship between PBI scores and 4-month outcomes after treatment with antidepressants was explored in 60 outpatients with major depression, controlling for potentially confounding factors. Results: A multiple logistic regression analysis suggested that low levels of paternal care, unmarried condition, non-melancholic features, and a high isolation tendency were all factors that contributed to poor outcomes for depression. The predictive power of low paternal care was not influenced by levels of depression or neuroticism. Limitation: This study did not attempt to explore whether the effects of parenting of father and mother on outcomes for depression may differ between male and female subjects. Conclusion: The results suggest that low levels of paternal care may be an independent predictor of a poor response to treatment with adequate antidepressants. 1999 Elsevier Science B.V. All rights reserved. Keywords: Parental bonding instrument; Major depression; Treatment outcome; Personality
1. Introduction Theorists and clinicians have long assumed that exposure in childhood to dysfunctional parenting is a
*Corresponding author. Tel.: 1 81-25-2272213; fax: 1 81-252270777. E-mail address: [email protected] (K. Sakado)
risk factor for depression in adulthood (see Becker, 1974; Parker, 1983a). Subsequent research in this area has led to a proposal of two principal factors in perceived child–parent relationships which may be relevant to depression in adulthood: lack of warmth or care and negative control (overprotection) (Parker et al., 1979; Raskin et al., 1971). Since the first report by Parker (1979), a large number of casecontrol and community studies (Kerver et al., 1992;
0165-0327 / 99 / $ – see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S0165-0327( 98 )00062-7
60
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
Mackinnon et al., 1993; Parker, 1983b; Parker et al., 1995; Parker, 1992; Plantes et al., 1988; Rodgers, 1996; Sato et al., 1997) have empirically confirmed that these two factors, particularly when measured by the Parental Bonding Instrument (PBI), can be associated with the development of depression or a depressive state in adulthood (particularly nonmelancholic or reactive depression). It has been suggested that the PBI can be correlated with not only onset, but also the course or outcome of depression. Gotlib et al. (1988) followed depressed postpartum women for an average of 30 months, and found that subjects who had received a lower level of parental care were still symptomatic at the last follow-up. They therefore have concluded that the level of perceived care is predictive of levels of depression later in life. However, there are some limitations in these findings. First, because the subjects in their study were comprised of depressed postpartum women as judged by the current level of the Beck Depression Inventory (Beck et al., 1961), it is questionable whether similar findings will be found in subjects diagnosed as having a depressive disorder. Second, it is unclear whether the subjects received treatment during the follow-up period, and if they had, it is not known how they were treated. In addition, the study did not examine the potentially confounding effects of other psychosocial factors (such as personality pathology), which are wellknown to influence the outcome of depression. Unfortunately, no systematic study has investigated the relationship between perceived parental characteristics and the outcome of depression, if the limitations of the Gotlib et al. (1988) study are considered carefully. The present study investigates the influence of parenting behaviors and attitudes, as measured by the PBI, on 4-month outcomes for depression. To minimize the limitations noted in the study described above, this study was designed as follows: we chose outpatients diagnosed as having major depression according to the Diagnostic and Statistical Manual of Mental Disorders, 3rd edition, revised (DSM-III-R). The subjects were followed closely while receiving adequate doses of antidepressants. Logistic regression analysis was used to separate out the effects of personality pathologies as well as demographic or
clinical characteristics on the outcomes for depression.
2. Subjects and methods
2.1. Materials and procedure The sample group consisted of new patients who were consecutively admitted to the outpatient clinic of the Psychiatric Department at Niigata City General Hospital from April 1994 to January 1995. After informed consent was obtained, the Structured Clinical Interview for DSM-III-R (SCID; Spitzer et al., 1990) was administered to depressed patients within 1 week of their first clinical examination. The patients diagnosed with major depression by SCID were included in the study. Patients with psychotic features, a lifetime history of hypomanic or manic episodes, panic or generalized anxiety disorders, alcohol or substance abuse, or organic brain syndrome were excluded. Patients younger than age 21 or older than age 65 were also excluded. A total of 69 patients met the criteria for initial inclusion. The patients were prescribed adequate levels of tricyclic antidepressants for at least 20 weeks. Either imipramine, clomipramine, amitriptyline, or maprotiline was used for each patient, with small doses of benzodiazepines and / or levomepromazine. The efficacy of maprotiline is believed to be equal to that of tricyclic antidepressants (Logue et al., 1979). If there was an inadequate response to the antidepressants, dosages were increased every week during the initial 8 weeks up to the maximum allowed by the Japan Pharmaceutical Information Center (1995) (200mg / day for imipramine; 225mg / day for clomipramine; 150mg / day for amitriptyline; 200mg / day for maprotiline). If a patient’s response to the medication was still inadequate, the maximum dosage was used for the remainder of the study (at least 12 weeks). During the follow-up period, six patients dropped out. Two patients who required inpatient treatment were referred to other hospitals, one patient underwent a manic episode, and three patients discontinued treatment. In addition, three patients were excluded because they could rate only one parent on
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
the PBI scale, due to the death of the other parent. The remaining 60 patients were included in the study. Of the 60 patients, 30 (50%) were male and 30 (50%) were female. The mean age was 43.0 (s.d. 5 12.0; range 22–64) years. Thirteen patients (22%) were unmarried and 47 patients (78%) were married. The mean of education was 12.0 (s.d. 5 2.4; range 8–16) years. Ten patients (17%) had no regular occupation. The mean score on the 17-item Hamilton Rating Scale for Depression (HAM-D; Hamilton, 1960) at baseline was 19.2 (s.d. 5 5.8; range 12–38). The mean age of onset for the first depressive episode was 41.0 (s.d. 5 12.6; range 16– 63) years. The number of patients with recurrent major depression was 18 (30%). The depressive state for 35 (58%) of the patients met the criteria for the melancholic subtype, according to the DSM-III-R. Of the 60 patients, 11 (18%) had a family history of affective disorders. There was no significant difference between the 60 patients included in this study and the 9 patients excluded during the follow-up period in terms of the sex ratio (% of female: 50 v. 44%), mean age (43.0 v. 43.6 years), marital status (% of unmarried: 22 v. 22%), mean education (12.0 v. 13.3 years), occupation (% of no regular occupation: 17 v. 11%), HAM-D score at baseline (19.2 v. 17.6), mean onset age of first depressive episode (41.0 v. 43.0), number of patients with recurrent major depression (30 v. 22%), melancholic features (% of melancholic subject: 58 v. 67%), or family history of affective disorders (% of affective disorders: 18 v. 22%). The severity of depression of the patients was evaluated three different times using the HAM-D: at the baseline, after 8 weeks, and after 16 weeks. At the second assessment of depressive symptoms, the subjects also completed the PBI and MPT (von Zerssen et al., 1988), which is a short-item selfrating questionnaire assessing six personality dimensions (see below). With respect to the MPT, subjects were asked to respond to the questionnaire based on their assessment of themselves in a physically and mentally healthy state. The 4-month outcome (remitted or non-remitted) was determined for each patient according to the following criteria of remission: (1) less than eight points on the HAM-D, as proposed by Frank et al.
61
(1991); (2) a full recovery in social function at the 16-week follow-up; and (3) no sign of recurrence of depression during the 4 weeks after the assessment according the two criteria listed above. Outcomes were assessed by one of us (K.S. or T.U.), both of us being blind to both the clinical data at baseline and scores on the PBI and MPT.
2.2. Measures The PBI is a self-report questionnaire developed by Parker et al. (1979). Respondents answer 25 questions about each parent, based on their memories of parents during the first 16 years of their childhood. Using this questionnaire, two child-rearing styles (care and protection) can be assessed for each parent. These parenting styles are derived by factor analysis. In addition, it has been suggested that the PBI can measure with validity actual, and not merely perceived, parental behavior (Parker, 1992). Kitamura and Suzuki (1993) has developed the Japanese version of the PBI using the back-translation method. Their validation study has found that the factor structure of the Japanese version is similar to that of the original questionnaire. In the present study, we used the Japanese version of the questionnaire to assess parenting styles. To assess the personality of the subjects, the MPT (Munich Personality Test; von Zerssen et al., 1988) was used. The MPT is a self-rating questionnaire of 51 items designed to assess six personality dimensions: Neuroticism, Extraversion, Frustration Tolerance, Rigidity, Isolation Tendency, and Esoteric Tendencies. Neuroticism and Extraversion in the MPT are nearly identical to the corresponding concepts advocated by Eysenck (1959). Frustration Tolerance primarily measures traits like emotional stability and insensitivity. Rigidity represents a kind of obsessionality (love of orderliness, dependence on norms and perfectionism, and so on). Both the Isolation Tendency and Esoteric Tendencies describe personality pathologies similar to schizoid and schizotypal personality disorders defined in the DSM-III-R. We used the Japanese version of the MPT, which was translated in keeping with the back-translation method, and has shown a high level
62
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
of consistency with the original test (Sakado et al., 1996).
2.3. Analysis To determine whether assessments of recollections of parenting behaviors and personality were independent of the level of depression, correlation of the HAM-D score at the 8-week follow-up, when the PBI and MPT were administered, to the PBI and MPT dimensional scores was calculated prior to other analyses. The sample was divided into remitted and nonremitted patients on the basis of the 4-month outcome assessment. The demographic and clinical data were univariately compared between the two groups. Similarly, mean scores on each dimension of both the PBI and the MPT were compared between the two groups. In addition, to examine whether or how these variables contributed to the outcome, a multiple logistic regression procedure was performed, using any variables which reached a significant level in the above analyses. To obtain the most appropriate model, Akaike’s information criterion (AIC; Akaike, 1987) was calculated, and a combination of the variables producing the minimum AIC was selected. For the analyses, a computer package, ‘‘SPSS for Windows 6.1J’’ was used. The significant level was set at 0.05 throughout this study.
3. Results
3.1. Correlation of MPT and PBI dimensional scores to depressive symptoms The mean HAM-D score at the 8-week follow-up, when the PBI and MPT were administered, was 3.6 (s.d. 5 3.4), suggesting that assessments of recollection of parenting behaviors and personality were independent of depressive level. This was confirmed by calculating the correlation coefficients of the HAM-D score at the 8-week follow-up to the PBI and MPT dimensional scores. Coefficients were: 2 0.16 (P 5 0.23) for paternal care, 0.18 (P 5 0.17) for paternal protection, 0.14 (P 5 0.30) for maternal
care, 0.02 (P 5 0.89) for maternal protection, 0.19 (P 5 0.18) for Extraversion, 0.16 (P 5 0.22) for Neuroticism, 0.03 (P 5 0.80) for Frustration Tolerance, 2 0.01 (P 5 0.91) for Rigidity, 0.03 (P 5 0.84) for Isolation Tendency, and 0.19 (P 5 0.18) for Esoteric Tendencies. No coefficients were significant. There was a high and significant correlation between paternal and maternal care scores (r 5 0.48, P 5 0.000) and between paternal protection and maternal protection scores (r 5 0.65, P 5 0.000) in the sample.
3.2. Univariate analysis Thirty six patients (60%) were judged to have remitted and 24 (40%) had not remitted at the 4-month follow-up. The demographic and clinical characteristics are shown in Table 1, together with the comparison between the remitted and non-remitted groups. In this comparison, some significant differences were found between the two groups. The mean age in the remitted group was significantly higher than in the non-remitted group (t 5 2 2.7, d.f. 5 58, P 5 0.009). The ratio of patients unmarried in the remitted group was significantly lower than in the non-remitted group ( x 2 5 5.9, d.f. 5 1, P 5 0.02). The ratio of patients with melancholic features in the remitted group was significantly higher than in the non-remitted group ( x 2 5 7.1, P 5 0.008). There were no other significant differences in demographic and clinical variables between the two groups. Table 2 shows the scores on each PBI or MPT dimension for both the remitted and non-remitted groups. A comparison between the two groups is also given. On the PBI, the mean score for paternal care in the remitted group was significantly higher than in the non-remitted group (t 5 2 2.2, d.f. 5 58, P 5 0.03). On the MPT, the remitted group showed significantly lower mean scores for Extraversion (t 5 2 2.3, d.f. 5 58, P 5 0.03), Neuroticism (t 5 2 2.8, d.f. 5 58, P 5 0.008), Isolation Tendency (t 5 2 2.8, d.f. 5 58, P 5 0.008), and Esoteric Tendencies (t 5 2 2.5, d.f. 5 34.9, P 5 0.02) than did the non-remitted group.
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
63
Table 1 Demographic and clinical characteristics of the subjects Variables
Sex ratio, female, n (%) Age, mean (s.d.), years Education, mean (s.d.), years Marital status, n (%) Married Unmarried No regular occupation, n (%) HAM-D, mean (s.d.) Onset, mean (s.d.), years Number of patients with recurrent major depression, n (%) Melancholic features according to DSM-III-R, n (%) Family history of major affective disorders, yes, n (%)
All subjects
Remitted
Non-remitted
Comparison between groups
(n 5 60)
(n 5 36)
(n 5 24)
t or x 2
d.f.
P
95% CI
x 5 0.0 t 5 2.7 t 5 2 1.3 x 2 5 5.9
1 58 58 1
NS (1-tailed) P 5 0.009 (2-tailed) NS (2-tailed) P 5 0.02 (1-tailed)
2 0.3 to 0.3 2.2 to 14.2 2 2.1 to 0.4 0.04 to 0.5
2 0.1 to 0.3 2 2.8 to 3.6 2 0.3 to 12.7 2 0.2 to 1.72
2
30 (50) 43.7 (13.8) 12.0 (2.4)
18 (50) 46.3 (10.8) 11.7 (2.4)
12 (50) 38.1 (12.3) 12.5 (2.3)
47 (78) 13 (22) 10 (17) 19.2 (5.8) 41.0 (12.6) 18 (30)
32 (89) 4 (11) 4 (11) 19.4 (6.1) 43.6 (12.7) 11 (31)
15 (63) 9 (38) 6 (25) 19.0 (6.0) 37.4 (29) 7 (29)
Fisher’s exact test t 5 0.2 t 5 1.9 x 2 5 0.01
58 58
NS NS NS NS
35 (30)
26 (31)
9 (38)
x 2 5 7.1
1
P 5 0.008 (1-tailed)
0.1 to 0.6
11 (18)
9 (25)
2 (8)
x 2 5 2.7
1
NS (1-tailed)
2 0.01 to 0.3
(2-tailed) (2-tailed) (2-tailed) (1-tailed)
Table 2 Comparison of PBI and MPT between remitted and not remitted groups
PBI Paternal care Paternal protection Maternal care Maternal protection MPT Extraversion Neuroticism Frustration tolerance Rigidity Isolation tendency Esoteric tendencies
(A)
(B)
Comparison between groups
Remitted (n 5 36) mean (s.d.)
Non-remitted (n 5 24) mean (s.d.)
t
d.f.
P (2-tailed)
95% CI
23.7 11.2 23.7 11.2
(5.9) (6.4) (6.4) (6.8)
19.8 12.3 24.3 11.1
(7.8) (6.1) (5.8) (8.0)
2.2 2 0.6 2 0.4 0.0
58 58 58 58
0.03 NS NS NS
0.3 to 7.4 2 4.3 to 2.3 2 3.9 to 2.6 2 3.8 to 3.9
A.B
8.0 10.1 5.0 11.1 3.6 1.7
(4.5) (4.7) (3.3) (4.3) (2.0) (1.6)
10.7 14.1 4.7 12.0 5.2 3.1
(4.4) (6.5) (3.7) (5.4) (2.7) (2.5)
2 2.3 2 2.8 0.3 2 0.7 2 2.8 2 2.5
58 58 58 58 58 34.9
0.03 0.008 NS NS 0.008 0.02
2 5.0 to 2 6.9 to 2 1.5 to 2 3.4 to 2 2.9 to 2 2.6 to
A,B A,B
2 0.3 2 1.1 2.1 1.6 2 0.4 2 0.3
A,B A,B
PBI, Parental Bonding Instrument; MPT, Munich Personality Test; CI, confidence interval.
3.3. Multiple logistic regression analysis The univariate analyses revealed that three of the demographic and clinical variables (age, marital status, and melancholic features), one of the PBI dimensional scores (paternal care), and four of the MPT dimensional scores (Extraversion, Neuroticism, Isolation Tendency, and Esoteric Tendencies) may have been predictive of the 4-month outcome for the
subjects. These eight variables were therefore considered as possible candidates for a logistic model that could best fit the data. Although a stepwise procedure is often used to identify subsets of independent variables that are good predictors of the dependent variable, there is a possibility that this procedure may not yield a ‘‘best’’ variable set. Therefore, all combinations of those eight variables were explored by performing 255 (2 8 2 1) logistic
64
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
regression analyses with the 4-month outcome as the dependent variable, and comparisons of model fits were made by calculating AIC for each logistic model. AIC is known to reach a minimum when the proposed model best fits the data (Akaike, 1987). The AIC calculations in this study demonstrate that a combination of four variables (marital status, melancholic features, paternal care of the PBI, and Isolation Tendency of the MPT) gave the minimum AIC value, indicating that the logistic model including these four variables was the most appropriate in predicting the 4-month outcome. The equation for this logistic model is shown in Table 3. The AIC for the model was 65.5, which was significantly lower than the null model including only a constant (AIC 5 80.8; x 2 5 25.2, d.f. 5 4, P 5 0.0000). In the equation, all four variables are significant in predicting the 4-month outcome. The odds ratios for having a poorer treatment outcome were: 6.0 for marital status (unmarried subjects against married subjects), 6.9 for melancholic features (subjects with non-melancholic features against those with melancholic features), 1.9 for paternal care (subjects reporting a five-point lower score), and 2.5 for Isolation Tendency (subjects reporting a three-point higher score), and all were significant. The odds ratios were judged to be significant if their 95% confidence intervals (CIs) did not include 1.0. It has been suggested that PBI scores may be influenced by response biases caused by the level of depression or trait characteristics such as neuroticism (Parker, 1983b). Correlation analyses in this study, however, show that the PBI scores are independent of the level of depression. We also attempted to
determine whether neuroticism has an influence on the relationship between low paternal care and poor treatment outcome indicated by the above logistic regression analysis. The Neuroticism score from the MPT was entered into the best-fitting model described above, and the predictive values of the variables included in the equation were observed. Using the Neuroticism score, the model gave an AIC value of 67.4, indicating that this model is a poorer fit for the data than that described in Table 3. The odds ratios for having a poorer treatment outcome were: 5.0 (NS) for marital status (unmarried subjects against married subjects), 6.7 (P , 0.05, 95% CI 5 1.6–27.7) for melancholic features (subjects with non-melancholic features against those with melancholic features), 1.9 (P , 0.05, 95% CI 5 1.7–2.2) for paternal care (subjects reporting a five-point lower score), 2.3 (NS) for Isolation Tendency (subjects reporting a three-point higher score), and 1.1 (NS) for Neuroticism (subjects reporting a threepoint higher score). Probably because there is a high correlation between Neuroticism and marital status (r 5 0.47, P 5 0.000 when treating marital status as a dummy variable), and between Neuroticism and Isolation Tendency (r 5 0.41, P 5 0.001), these two variables lose much of their predictive power in the equation. However, non-melancholic features and paternal care remained significant.
4. Discussion The present study attempted to explore whether parenting behaviors, as measured by the PBI, in-
Table 3 Results of multiple logistic regression analysis Independent variables Dependent variable 0 5 remitted 1 5 non-remitted AIC 5 65.5 Model R 2 5 0.46 Unmarried Non-melancholic features Paternal care Isolation Tendency a
b
Odds ratio
95% CI
Wald statistics
1.8 1.9 2 0.1 0.3
6.0 a 6.9 a 1.9 (for a 5-point score lower)a (2.5 (for a 3-point score higher)a
1.2 to 30.3 3.4 to 13.8 1.7 to 2.2 1.9 to 3.4
4.7 (P 5 0.03) 7.1 (P 5 0.008) 5.2 (0.02) 3.9 (0.048)
Significant odds ratio. Subjects reporting a 5-point lower score for paternal care are 1.9 times more likely to have a poorer outcome. AIC, Akaike’s Information Criterion; CI, confidence interval. b
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
fluence short-term outcomes for depressive outpatients treated with adequate antidepressants. Patients with poorer treatment outcomes reported a significantly lower score for paternal care than did those who responded well to the antidepressants, and this difference remained significant even when the possible effects of demographic, clinical, and personality variables were separated out. The multiple logistic regression analyses conducted in this study demonstrate that subjects reporting a five-point lower score for paternal care were 1.9 times more likely to have a poorer response to treatment with adequate antidepressants. The correlation analyses conducted for this study indicate that the PBI scores were independent of the level of depression. In the logistic model which includes Neuroticism, paternal care remains a significant predictor of response to antidepressants. These results indicate that the association between low paternal care and a poor response to antidepressants are not due to response biases caused by the level of depression or trait characteristics such as neuroticism, suggesting that there may be a causal linkage. One might expect that a low score for paternal care would be associated with non-melancholic features of depression, as previous studies on the PBI in depressive patients have suggested (Parker, 1992; Parker and Hadzi-Pavlovic, 1993). This association, it would seem, could account for the significance of low paternal care in predicting a poor response to antidepressants. However, the most appropriate logistic model including both paternal care and nonmelancholic features shows that paternal care is significant in predicting the poor response even after the effects of the non-melancholic features are separated out. The results of this study suggest that low paternal care is a significant risk factor for a poor response to treatment with adequate antidepressants. This risk appears to be independent of nonmelancholic features, personality assessments, and other demographic variables. Although Gotlib et al. (1988) were the first to find that low parental care may contribute to poor outcomes for depression, some methodological limitations in their study, as mentioned above, make it difficult to interpret their findings. The present study, which used a more clearly defined sample and was better controlled for treatment conditions and other
65
potentially confounding factors, provides more direct evidence for an independent link between low parental care and outcomes for depression. The results of this study may enhance researchers’ interest in the association between parenting behaviors and depression. Most previous studies in this area have explored parenting behavior as a cause of depressive disorders in adulthood. Evidence derived from these studies suggests that low parental care rather than high parental protection may be a primary risk factor for depressive disorders (Mackinnon et al., 1993; Parker, 1983a). It is still controversial, however, whether paternal or maternal care is more influential in the development of depressive disorders (Rodgers, 1996). This study indicates that only low paternal care (but not maternal care) is significant in predicting poor treatment outcomes in patients with major depressive disorder, although there does appear to be a high correlation between paternal and maternal care scores. These results suggest that low paternal care may be relevant not only to the pathogenesis of depressive disorders but also to poor treatment outcomes for depression, and that the influence on these outcomes may differ when considering levels of paternal and maternal care. Similar studies are needed to replicate and confirm the findings of this study. Although this study found no sex difference in the outcomes between the remitted and non-remitted groups, there is a possibility that the effects of parenting of father and mother on outcomes for depression may be different between male and female subjects. Unfortunately, we could not analyze the data after separating the subjects into men and women, because our sample size was somewhat small. Naturally, a further study is needed to explore whether the relationship of the PBI to outcomes for depression differs between male and female subjects. Based on the results of this study, it is unclear how low paternal care is causally linked to a poor response to antidepressants. Parker et al. (1992) have reviewed previous studies investigating the continuity between parenting behavior, as measured by the PBI, and social bonds in adulthood. They have pointed out that there is some evidence for a link between low parental care and deficiencies in diffuse social bonds. In addition, some studies have suggested that depressive patients with deficiencies in
66
K. Sakado et al. / Journal of Affective Disorders 52 (1999) 59 – 66
social bonds have a poorer response to antidepressants (Lam et al., 1994). One can speculate that deficiencies in social bonds may be responsible for the causal linkage between low parental care and a poor response to antidepressants. On the other hand, considering that Kendler (1996) found evidence of the effect of genetic factors on parental care, a poor response to antidepressants in subjects who received low parental care may be influenced by genetic factors.
References Akaike, H., 1987. Factor analysis and AIC. Psychometrika 52, 317–332. Beck, A.T., Ward, C.H., Mendelson, M., et al., 1961. An inventory for measuring depression. Arch. Gen. Psychiatry 4, 561–571. Becker, J., 1974. Depression. Theory and Research. Winston, Washington, DC. Eysenck, H.J., 1959. Manual of the Maudsley Personality Inventory. University of London Press, London. Frank, E., Prien, R.F., Jarrett, R.B., et al., 1991. Conceptualization and rationale for consensus definitions of terms in major depressive disorder. Remission, recovery, relapse, and recurrence. Arch. Gen. Psychiatry 48, 851–855. Gotlib, I.H., Mount, J.H., Cordy, N.I., et al., 1988. Depression and perceptions of early parenting: a longitudinal investigation. Br. J. Psychiatry 152, 23–27. Hamilton, M., 1960. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23, 56–62. Japan Pharmaceutical Information Center, 1995. Drug in Japan: Ethical Drugs. Yakugyojiho-Sya, Tokyo. Kendler, K.S., 1996. Parenting: a genetic-epidemiological perspective. Am. J. Psychiatry 153, 11–20. Kerver, M.J., van Son, M.J.M., de Groot, P.A., 1992. Predicting symptoms of depression from reports of early parenting: a one-year prospective study in a community sample. Acta Psychiatr. Scand. 86, 267–272. Kitamura, T., Suzuki, T., 1993. A validation study of the Parenting Bonding Instrument in a Japanese population. Jap. J. Psychiatr. Neurol. 47, 29–36. Lam, D.H., Green, B., Power, M.J., et al., 1994. The impact of social cognitive variables on the initial level of depression and recovery. J. Affect. Disord. 32, 75–83. Logue, J.N., Sachais, B.A., Feigher, J.P., 1979. Comparisons of maprotiline with imiparamine in severe depression: a multicenter controlled trial. J. Clin. Pharmacology. 19, 64–74.
Mackinnon, A.J., Henderson, A.S., Andrew, G., 1993. Parental ‘‘affectionless control’’ as an antecedent to adult depression: a risk factor refined. Psychol. Med. 23, 135–141. Parker, G., Tupling, H., Brown, L.B., 1979. A Parental Bonding Instrument. Br. J. Med. Psychol. 59, 1–10. Parker, G., 1979. Parental characteristics in relation to depressive disorders. Br. J. Psychiatry 134, 138–147. Parker, G., 1983a. Parental Overprotection: A Risk Factor in Psychosocial Development. Grune and Stratton, New York, London, Paris, San Diego, San Francisco, Sao Paulo, Sydney, Tokyo and Toronto. Parker, G., 1983b. Parental ‘‘affectionless control’’ as an antecedent to adult depression. Arch. Gen. Psychiatry 40, 956–960. Parker, G., 1992. Early environment. In: Paykel, E.S. (Ed.), Handbook of Affective Disorders, 2nd ed. Churchill Livingstone, Edinburgh, London, Madrid, Melbourne and Tokyo. Parker, G., Barrett, E., Hickie, I.B., 1992. From nurture to network: examining links between perceptions of parenting received in childhood and social bonds in adulthood. Am. J. Psychiatry 149, 877–885. Parker, G., Hadzi-Pavlovic, D., 1993. Parental representation of melancholic and non-melancholic depression: examining for specificity to depressive type and for evidence of additive effects. Psychol. Med. 22, 657–665. Parker, G., Hadzi-Pavlovic, D., Greenwald, S., et al., 1995. Low parental care as a risk factor to lifetime depression in a community sample. J. Affect. Disord. 33, 173–180. Plantes, M.M., Prusoff, B.A., Brennan, J., et al., 1988. Parental representations of depressed outpatients from a U.S.A. sample. J. Affect. Disord. 15, 149–155. Raskin, A., Boothe, H.H., Reatig, N.A., et al., 1971. Factor analyses of normal and depression patients’ memories of parental behavior. Psychiatr. Rep. 29, 871–879. Rodgers, B., 1996. Reported parental behavior and adult affective symptoms. 1. Associations and moderating factors. Psychol. Med. 26, 51–61. Sakado, K., Sato, T., Uehara, T., et al., 1996. The Japanese version of the Munich Personality Test (MPT): its reliability and validity. Arch. Psychiatr. Diag. Clin. Eval. 7, 123–132. (in Japanese). Sato, T., Sakado, K., Uehara, T., et al., 1997. Perceived parental style in a Japanese sample of depressive disorders. A replication outside Western culture. Br. J. Psychiatry 170, 173–175. Spitzer, R., Williams, J., Gibbon, M. et al., 1990. Structured Clinical Interview for DSM-III-R (SCID). User’s Guide. APA, Washington, DC. von Zerssen, D., Pfister, H., Kleller, D.M., 1988. The Munich Personality Test (MPT). A short questionnaire for self-rating and relatives’ rating of personality traits: formal properties and clinical potential. Eur. Arch. Psychiatr. Neurol. Sci. 238, 73– 93.