Results
Associations between CA and social support in adulthood
Table
1 shows mean social support scores for each quintile of CA at each assessment period at 18, 21, 25, 30, 35 and 40 years. The table also provides a pooled estimate of the mean and standard deviation of social support for each level of CA over the assessment periods from age 18 to age 40, as well as a test of significance for the pooled association between CA and social support as derived from random-effects GEE models.
Table 1
Mean (sd) scores on the standardised social support score for each CA quintile at ages 18, 21, 25, 30, 35 and 40 years, pooled across observations
18 | 100.41 (10.10) | 100.44 (9.44) | 100.91 (9.93) | 98.56 (10.45) | 98.96 (10.82) |
21 | 101.70 (9.53) | 100.46 (9.85) | 100.36 (9.80) | 97.84 (10.20) | 97.59 (11.69) |
25 | 102.65 (9.67) | 100.58 (9.82) | 99.23 (9.43) | 96.61 (10.36) | 99.20 (10.72) |
30 | 102.10 (9.22) | 101.23 (8.38) | 99.79 (9.33) | 96.55 (11.10) | 95.91 (13.35) |
35 | 102.09 (8.84) | 101.35 (8.83) | 99.82 (8.82) | 96.90 (12.20 | 93.35 (9.87) |
40 | 102.44 (8.77) | 101.38 (8.88) | 99.00 (8.99) | 97.47 (12.13) | 91.64 (13.94) |
Pooled score | 101.88 (9.40) | 100.89 (9.22) | 99.87 (9.41) | 97.32 (11.05) | 96.36 (11.97) |
The table shows that there was a dose-dependent pattern in the association between increasing levels of CA exposure and lower levels of social support that was consistent across assessment periods. The pooled association was found to be statistically significant (B(SE) = − 1.40 (0.18), p < 0.001). Examination of the pooled scores show that those with the lowest level of CA exposure had the highest average social support score of 101.88 (sd = 9.40), compared to those with the highest exposure who had the lowest average social support score of 96.36 (sd = 96.36).
Associations between CA and internalising disorder in adulthood
Table
2 shows the frequency of major depression and any anxiety disorder for each quintile of exposure to CA, at each assessment period at 16–18, 18–21, 21–25, 25–30, 30–35 and 35–40 years. The table also provides a pooled estimate of the prevalence of major depression and any anxiety disorder for each level of CA over the assessment periods from 16–18 years to 35–40 years, as well as respective tests of significance of the pooled associations between CA and major depression and any anxiety disorder, after adjusting for covariates, as derived from random effects GEE models.
Table 2
Percent meeting criteria for major depression and anxiety disorder, respectively, for each CA quintile at ages 16–18, 18–21, 21–25, 25–30, 30–35 and 35–40, pooled across observations
16-18 | 11.0 | 13.3 | 26.6 | 35.0 | 65.0 |
18-21 | 14.1 | 21.1 | 27.8 | 31.2 | 53.8 |
21-25 | 13.8 | 16.9 | 24.1 | 35.0 | 48.7 |
25-30 | 12.9 | 19.5 | 24.2 | 26.8 | 44.7 |
30-35 | 12.1 | 14.7 | 17.7 | 28.9 | 52.8 |
35-40 | 16.5 | 15.7 | 20.0 | 33.6 | 58.3 |
ORs (95%CI) | 1 | 1.5 (1.41, 1.77) | 2.49 (1.98, 3.14) | 3.94 (2.78, 5.56) | 6.21 (3.92, 9.86) |
16-18 | 7.6 | 10.8 | 17.6 | 31.4 | 60.0 |
18-21 | 5.5 | 9.5 | 15.7 | 19.6 | 38.5 |
21-25 | 5.9 | 12.9 | 17.6 | 27.3 | 35.9 |
25-30 | 11.2 | 13.9 | 17.7 | 25.4 | 44.7 |
30-35 | 9.3 | 11.6 | 14.8 | 27.4 | 38.9 |
35-40 | 13.9 | 11.7 | 23.0 | 29.4 | 50.0 |
OR (95%CI) | 1 | 1.53 (1.36, 1.73) | 2.35 (1.84, 3.01) | 3.61 (2.50, 5.21) | 5.54 (3.40, 9.03) |
Table
2 shows that there was a dose-dependent association between increasing levels of CA and higher prevalence of both major depression and any anxiety disorder that was consistent across all time periods. The adjusted pooled associations were found to be statistically significant (both p < 0.001). Examination of the pooled prevalence shows that those with the highest level of CA exposure had odds of meeting criteria for major depression that were 6.21 times higher than those with the lowest level of exposure. The pooled prevalence also indicate that those with the highest level of CA exposure had odds of meeting criteria for any anxiety disorder that were 5.54 times higher than those with the lowest level of exposure. The strength of the association between CA exposure and major depression weakened over time (B = − 0.05, p = 0.01), as did the strength of the association between CA exposure and any anxiety disorder (B = − 0.06, p = 0.002).
CA * time period interaction term
In addition, there was a statistically significant interaction term for CA and time period for the models of both major depression and any anxiety disorder (major depression B(SE) = − 0.05 (0.02), p = 0.01); anxiety disorder B(SE) = − 0.06 (0.02) p = 0.002). In both cases, the interaction term suggested that the magnitude of the association between CA and both major depression and any anxiety disorder decreased over the assessment periods from ages 16–18 years to 35–40 years.
Associations between social support and internalising disorder
Table
3 shows the prevalence of major depression and any anxiety disorder (respectively) for each quintile of the social support score, at each assessment period at 16–18, 18–21, 21–25, 25–30, 30–35, and 35–40 years. The table also provides pooled estimates of the prevalence of major depression and any anxiety disorder, respectively, for each level of social support over the assessment periods from age 16–18 to age 35–40, adjusted for covariates, as derived from random effects GEE models.
Table 3
Percent meeting criteria for major depression and any anxiety disorder, respectively, for each social support quintile at ages 16–18, 18–21, 21–25, 25–30, 30–35 and 35–40, pooled across observations
16–18 | 26.4 | 21.7 | 23.5 | 20.7 | 14.8 |
18–21 | 28.7 | 28.1 | 21.9 | 15.5 | 18.1 |
21–25 | 34.8 | 27.3 | 25.0 | 17.4 | 11.6 |
25–30 | 31.0 | 18.8 | 21.6 | 17.0 | 17.4 |
30–35 | 29.0 | 15.1 | 16.0 | 14.2 | 18.4 |
35–40 | 33.3 | 20.8 | 22.0 | 16.3 | 13.9 |
OR (95% CI) | 1 | 0.89 (0.84, 0.95) | 0.80 (0.70,0.91) | 0.71 (0.59, 0.86) | 0.63 (0.49,0.82) |
16–18 | 23.3 | 18.3 | 14.9 | 19.2 | 8.2 |
18–21 | 15.6 | 15.6 | 12.0 | 9.0 | 8.6 |
21–25 | 28.3 | 17.7 | 16.4 | 13.0 | 10.1 |
25–30 | 24.5 | 13.8 | 17.6 | 15.2 | 14.7 |
30–35 | 21.7 | 14.0 | 13.0 | 14.7 | 11.5 |
35–40 | 31.5 | 18.0 | 20.4 | 16.3 | 13.3 |
OR (95% CI) | 1 | 0.98 (.97, .99) | 0.78 (.68, .90) | 0.69 (.56, .86) | 0.61 (.46, .82) |
The table shows that there was an association between increasing levels of social support and lower prevalence of both major depression and any anxiety disorder. The adjusted pooled associations were found to be statistically significant (both p < 0.001). Examination of the pooled prevalence show that those with the highest level of social support had prevalence of major depression that were 37% lower than those with the lowest level of social support. The pooled prevalence also show that those with the highest level of social support had prevalence of any anxiety disorder that were 39% lower than those with the lowest level of social support.
Table
4 shows the results of the three GEE models testing first the association between CA and major depression and any anxiety disorder, second these associations adjusted for confounders, and third the mediating contribution of social support. Table
4 displays odds ratios with 95% confidence intervals representing the odds of major depression and anxiety, respectively, for those with the highest level of CA exposure as compared to those with the lowest level of CA exposure. It also displays the results of tests of significance of the effects. The prevalence of major depression and any anxiety disorder both increase as a function of higher CA exposure, both before and after adjustment for potentially confounding variables (Major Depression: OR (95% CI) = 1.73 (1.56, 1.91) and 1.59 (1.42, 1.78) respectively, both p < 0.001; any Anxiety Disorder: OR (95% CI) = 1.87 (1.67, 2.09) and 1.65 (1.47, 1.84) respectively, both p < 0.001).
Table 4
Odds of internalising disorders (Major depression and any anxiety disorder) in a birth cohort associated with increasing levels of cumulative childhood adversity: crude (Model 1), adjusted for confounders (Model 2) and further adjusted for social support (model 3)
Exposures |
Cumulative childhood adversity | 1.73 (1.56, 1.91)*** | 1.59 (1.42, 1.78)*** | 1.55 (1.39, 1.74)*** |
Social Support | | | 0.98 (.97, .99)*** |
CA x Time Period | 0.96 (.92, .99)** | 0.95 (.91, .98)** | 0.95 (.90, .98)** |
Covariates |
Biological sex | | 1.99 (1.57, 2.53)*** | 1.97 (1.55, 2.50)*** |
Father’s education | | 1.29 (1.10, 1.52)** | 1.31 (1.12, 1.54)** |
Neuroticism | | 1.07 (1.04, 1.10)*** | 1.07 (1.04, 1.10)*** |
Novelty seeking | | 1.03 (1.00, 1.05)* | 1.03 (1.00, 1.05)* |
Exposures |
Cumulative childhood adversity | 1.87 (1.67, 2.09)*** | 1.65 (1.47, 1.84)*** | 1.60 (1.43, 1.79)*** |
Social Support | | | 0.98 (.97, .99)*** |
CA x Time Period | 0.95 (.91, .98)** | 0.94 (.90, .98)** | 0.94 (.90, .98)** |
Covariates |
Biological sex | | 2.15 (1.66, 2.80)*** | 2.13 (1.64, 2.76)*** |
Neuroticism | | 1.08 (1.04, 1.17)** | 1.08 (1.04, 1.16)*** |
To test whether social support mediated the adjusted association between CA and major depression and any anxiety disorder, Model 3 was extended to include the time dynamic measure of social support (18, 21, 25, 30, 35 and 40 years) to the terms included in Model 2. The inclusion of social support (Model 3) reduced the magnitude of the association between CA and both major depression, and any anxiety disorder, but these associations remained statistically significant (Major depression: OR (95% CI) = 1.55 (1.39, 1.74), p < 001; any anxiety disorder: OR (95% CI) = 1.60 (1.43, 1.79) p < 0.001). Social support had a significant, negative association with major depression (OR (95% CI) = 0.98 (0.97, 0.99), p < 001) and any anxiety disorder (OR (95% CI) = 0.98 (0.97, 0.99), p < 001) after adjusting for CA and other covariates. This pattern of results suggests that social support played a weak but detectable mediating role in the associations between CA and internalising disorder.
Discussion
The present study aimed to examine the role of social support as a protective factor between cumulative CA and internalising disorders in adulthood. This is the first study of which the authors are aware to examine this effect using prospectively measured cumulative CA in a general population cohort.
There was a dose-dependent association between an accumulation of exposures to CA and increased odds of meeting criteria for major depression and anxiety disorders in adulthood. also found that social support plays a statistically significant but relatively weak role in reducing the magnitude of the relationship between CA and both major depression and any anxiety disorder in adulthood. This finding is consistent with the wider literature detailing the protective role of social support for mental health, particularly in adolescence and young adulthood, and for those who have experienced adversity in childhood [
36,
55‐
57]. Therefore, the findings of the present study provide evidence to the idea that support from a network of friends and family may protect against the development of internalising disorder in those with histories of high levels of exposure CA [
29]
The strengths of the present study overcome limitations present in previous studies of the protective effect of social support. First, the GEE framework enabled the utilisation of repeated measures of social support and internalising disorders to capitalize upon the richness of the available data and increased power. Second, the present study included control for confounding by several factors, which was possible due to the measurement of a wide range of variables from birth in the CHDS [
43]. The variables included were measured contemporaneously to the variables that constituted the CA measure [
18], therefore controlling more accurately the potentially confounding effects. Third, prospective measurement of CA is a key strength of the present study. The only exception was for abuse variables, which were validated with repeated measurements (age 18 and 21) [
58]. Very few longitudinal studies of social support as a protective factor for internalising outcomes of CA have measured CA prospectively [
29]. The current study therefore addressed an important gap in the literature by avoiding a large amount of the recall bias inherent in previous studies.
Although this study used birth cohort data, which provides some of the most rigorous data within observational research [
59], the study has its limitations. The CHDS cohort was representative of the NZ population at the time of the cohort’s birth, which is now over four decades ago, and as population demographics change over time, the cohort necessarily becomes less representative. This study represents the sociocultural context in which it was conducted, specific to both place and time. Similarly, a limitation of the present study is the use of measures that are, largely superseded. The social support measure used in the later waves (age 30–40 years) is more precise than the measure used in the earlier waves (age 18–25 years) because it accounts for the different ways in which friends provide support, compared to a measure of the size of one’s social network. Unfortunately, this is a common limitation of longitudinal studies, as they rely on what was relevant at the time of measurement. This limitation is offset considerably by the benefits of having repeated measures data and control of within-subjects variability.
Future studies should consider the relative influence of different aspects of social support. This may include the timing, the source (e.g. from family, friends, spouse), and the nature of the support (quality versus quantity). Though this study analysed the contribution of social support with repeated measures over time, it was not able to examine the particular influence of social support at different ages. Research suggests that social support may be particularly beneficial at times of transition, such as the transition from high school into employment or tertiary education [
60]. Therefore, a potential avenue for future research is to examine whether the presence of social support is particularly protective at these times. Further, some studies point to the role of different sources of support, and that support from family may be most influential [
36,
61]. They also suggest distinguishing quality and quantity of social support [
37,
61], though the effect may be strongest when both are enhanced [
56]. Examination of these distinctions with prospective cohort data could elucidate further details of how social support promotes resilience.
The findings of the present study provide empirical support for interventions based upon enhancing social support to improve internalising outcomes. This finding does, however, suggest that in common practice and discourse, the effect of social support may be overestimated. The small, though detectable, effect size observed indicates that social support alone is unlikely to exert enough difference to be clinically important. One explanation for the comparably convincing impact of social-support-based interventions is that enhancing social support has effects beyond increasing the size of the social network. For example, successful interventions may also drive self-esteem building, and psychoeducation [
62]. The combined effect of these factors requires investigation. The lack of strong effect sizes in this area of research suggests that when directing intervention efforts, the prevention of exposure to CA in the first place remains key.
This paper provides evidence that social support plays a weak but detectable role in mitigating major depression and anxiety disorders following CA. It suggests that while CA has a concerning association with adult internalising disorders, it is not deterministic. Social support may be one driver of positive outcomes following CA, but it may not be enough on its own. On a population level these findings do not suggest that there would be many detectable improvements in mental illness outcomes if social support alone were enhanced. The findings of the present study suggest that social support needs to be researched further to determine how it acts alongside other factors to improve internalising disorder outcomes. Social support is a key target for strategies that will reduce the long-term burden of CA, but this study also highlights the fundamental importance of preventing CA.