Weighing psychosocial factors in relatives for the risk of psychopathology: a study of patients with depressive and anxiety disorders and their siblings

One of the strongest risk factors for the onset of depressive and anxiety disorders is a family history of these disorders [1, 2]. A two- to three-fold increased risk of the disorders is found for siblings of probands with depressive and anxiety disorders as compared to individuals without affected relatives [3‐5]. However, despite their increased risk for psychopathology, siblings from the same at-risk family can differ substantially from one another in psychological functioning [6, 7]. Consistent with this, we showed in a previous study that proband-sibling resemblance in several psychopathology-related features (i.e. symptoms, social/cognitive vulnerabilities, personality traits) was only mild to moderate [8]. Although a large body of evidence exists for the association between several psychosocial risk factors (e.g. female gender, socioeconomic deprivation, social isolation, poor parental bonding, adverse events, smoking, alcohol (ab)use, physical inactivity) and high risk for depressive and anxiety disorders [9‐21], little is known about how these risk factors operate within families to increase psychopathology for at-risk siblings. Siblings’ increased risk for psychopathology may depend on the presence of such risk factors in their affected proband, either by also being present in the proband (e.g. an additional ‘vicarious’ effect) [22] or by being absent in the proband while being present in the sibling (e.g. the feeling of ‘being the black sheep’ in the family) [23‐25]. This may even extend to sociodemographic risk factors such as higher age and female gender: for instance, rumination seems to be ‘contagious’ especially among older and same-sex female sibling pairs potentially due to the stronger emotional bonds and social learning/sharing [26‐29]. Co-rumination, in turn, has been found to be associated with affective psychopathology [30, 31]. Identifying how psychosocial risk factors of poor mental health in siblings of affected probands operate within families may help identifying potential mechanisms explaining why some siblings develop a depressive and anxiety disorder, whereas others do not.

While a large number of studies have investigated the familial aggregation of depressive and anxious psychopathology, only a few have examined the impact of not only considering psychosocial risk factors within at-risk siblings, but also within their affected proband, on increasing psychopathology in at-risk siblings. Findings were mixed as to whether the individual risk for psychopathology is increased if risk factors in a relative are also present or if risk factors in a relative are absent. A larger neighborhood socioeconomic deprivation [32], higher childhood emotional maltreatment [33], and poorer parental bonding [34‐37] in an individual as compared to their sibling(s) was found to be associated with more severe depressive symptoms of that individual. Results were mixed for age and gender, with some studies finding associations of similarity (vs. dissimilarity) in female gender and age with similarity in depressive/anxious psychopathology [4, 38], while other studies found no added impact of taking into account the gender/age of an individual’s sibling for their risk for psychopathology [5, 39, 40]. Moreover, available studies have mainly been limited to investigate the degree but not the direction of proband-sibling (dis)similarity of risk factors [4, 5, 38, 40, 41] and mainly focused on sociodemographic and early life adversity risk factors, but not on a wider variety of psychosocial risk factors (e.g. also including recent life adversity and lifestyle-related factors).

The present study examined how a broad range of established psychosocial risk factors for depression/anxiety operate within families to explain interindividual differences in psychopathology between siblings of probands with depressive/anxiety disorders. The main aim was to disentangle and quantify the effect of the presence of a risk factor in the proband, by testing whether this (i) had a unique contribution for psychopathology in the sibling, over-and-above the presence of this risk factor in the sibling, and/or (ii) modified the association between this risk factor and psychopathology in the sibling.

The mean age of the sample (N = 636) was 49.7 years (SD = 13.2, range 20–78), mean years of education was 13.3, and 62.4% were female. Sample characteristics for probands and siblings separately can be found in Table 1. Of the 380 siblings, 191 (50.3%) had a lifetime depressive and/or anxiety disorder diagnosis. Missing data on study variables was small (Supplementary Table 1). Pairwise multilevel correlations between psychosocial risk factor variables can be found in Supplementary Table 2.

Associations of explanatory variables in siblings and probands and outcomes in siblings are reported in Table 2 (current symptom severity; linear mixed models) and Table 3 (lifetime psychiatric diagnosis; logistic mixed models). Analyses with individual-level sibling risk factors only (Step 1; Table 2) showed that more severe symptoms in the sibling were associated with female gender (γ = 0.33, SE = 0.08, p < 0.001), low income (γ = 0.39, SE = 0.08, p < 0.001), unemployment (γ = 0.32, SE = 0.09, p < 0.001), being single (γ = 0.24, SE = 0.10, p = 0.020), smoking (γ = 0.21, SE = 0.10, p = 0.030), hazardous alcohol use (γ = 0.26, SE = 0.09, p = 0.006), higher levels of childhood trauma (γ = 0.34, SE = 0.04, p < 0.001), and poorer maternal (γ = 0.25, SE = 0.04, p < 0.001) and paternal bonding (γ = 0.27, SE = 0.04, p < 0.001). These risk factors were also significantly associated with an increased likelihood of lifetime psychiatric diagnosis in the sibling (all p < 0.05; Step 1; Table 3), except for unemployment (OR = 1.64, SE = 0.42, p = 0.051).

In Step 2, we added the same risk factors assessed in probands as additional explanatory variables. For the outcome of current symptomatology in sibling, main effects of several proband individual-level risk factors were found (Step 2; Table 2): on top of the presence/degree of the risk factor in the sibling, proband low income (γ = 0.17, SE = 0.08, p = 0.045, ΔR²_marginal = 0.4%, ΔR²_conditional = 0.7%), being single (γ = 0.27, SE = 0.10, p = 0.008, ΔR²_marginal = 1.8%, ΔR²_conditional = 3.4%), and lower levels of childhood trauma (γ = − 0.10, SE = 0.04, p = 0.023, ΔR²_marginal = 0.2%, ΔR²_conditional = 2.7%), and more optimal maternal (γ = − 0.12, SE = 0.05, p = 0.013, ΔR²_marginal = 1.5%, ΔR²_conditional = 1.5%) and paternal bonding (γ = − 0.14, SE = 0.05, p = 0.003, ΔR²_marginal = 2.1%, ΔR²_conditional = 1.2%) were associated with more severe symptoms in the sibling. No significant main effects of proband risk factors on the outcome of lifetime psychiatric diagnosis in sibling were found (all p > 0.05; Step 2; Table 3).

The additional effect of sibling × proband interactions in risk factors was tested in Step 3. For the additive interaction effect on current symptomatology in the sibling (Step 3; Table 2), significant sibling × proband interactions were found for low education (γ = − 0.01, SE = 0.004, p = 0.028, ΔR²_marginal = 1.3%, ΔR²_conditional = 1.1%), childhood trauma (γ = − 0.001, SE = 0.0004, p = 0.006, ΔR²_marginal = 1.8%, ΔR²_conditional = 0.7%), and hazardous alcohol use (γ = − 0.53, SE = 0.19, p = 0.005, ΔR²_marginal = 2.2%, ΔR²_conditional = 2.4%). Figure 1 shows the association between a risk factor and symptoms in the sibling for different values of that risk factor in the proband for low education (left panel), childhood trauma (middle panel), and hazardous alcohol use (right panel). Consistently, when the risk factor was also present in the proband (lower years of education, higher levels of childhood trauma, and hazardous alcohol use), the strength of the association between the same risk factor and symptoms in their sibling was reduced. No significant (multiplicative) sibling × proband interactions were found for the outcome of lifetime psychiatric diagnosis in sibling (all p > 0.05; Step 3; Table 3), nor when the coefficient of the interaction term estimated departure from additivity (Supplementary Table 3).

In siblings of probands with lifetime depressive and anxiety disorders, we confirmed the association of a wide range of established psychosocial risk factors with psychopathology symptoms and disorders. However, the major finding of our study is that the presence of the same risk factors in affected probands, explained additional interindividual differences in psychopathology in at-risk siblings. For instance, siblings having low income and who were single had higher symptoms; intriguingly, the presence of the same risk factor in the proband was additionally associated with higher symptoms in their sibling, independently of the sibling’s individual risk factor. Furthermore, for other risk factors (low education, childhood trauma, hazardous alcohol use), the presence in the proband moderated the association between the risk factor and symptoms in the sibling: when the risk factor was also present in the proband, the strength of the association between the risk factor and symptoms in the sibling was reduced. Thus, when similar levels of a risk factor were shared between probands and siblings, the impact of the risk factor on siblings’ symptoms was buffered. Of note, we only confirmed these additional family effects of risk factors for the continuous outcome of psychopathology symptoms but not for binary clinical diagnoses, possibly due to the reduced statistical power or loss of information when using a dichotomous classification.

Besides showing an additive impact of probands’ low income, which is in line with findings from a previous community-based twin study [32], we extend the current literature by showing that when a proband was single, on top of the presence of the risk in their sibling, this was associated with more severe symptomatology in the sibling. This additive impact of also having low income and being single present in the proband may arise from additional familial clustering between these risk factors and psychopathology (i.e. a more genetic form of the disorders that tends to be co-inherited with these risk factors). In line with this reasoning, twin and genome-wide association studies have shown substantial genetic correlations of income (− 0.30 to − 0.44) [62, 63] and avoidant/anxious romantic attachment (0.48–0.58) [64], which might increase the probability of being single, with depressive and anxiety disorders. The additive impact of the presence of a risk factor in the proband may also be a reflection of the degree of (and/or other additional) problems within their family. That is, both proband and sibling having low income and being single may indicate a more substantial degree of a family’s socioeconomic deprivation and romantic relationship problems, which likely results from or leads to more severe affective problems [9, 65].

We also found several risk factors for which the additional presence or higher levels in the proband were associated with less severe symptomatology in their sibling(s). These factors included poor parental bonding and childhood trauma, which were also identified in previous studies [33‐37, 41], and low education and hazardous alcohol use. Independently of parental bonding levels in the sibling, poorer parental bonding in the proband was associated with less severe symptomatology in the sibling. For childhood trauma, low education, and hazardous alcohol use, the individual-level risk in the proband moderated the impact of the risk factor on symptomatology in their sibling(s): when these factors were similarly present in both proband and sibling, the impact of the factor on sibling symptomatology appeared to be buffered. Conversely, a sibling’s symptoms were higher when these risk factors were not shared with or were of lower level in their proband. This may reflect a ‘black sheep effect’, in which the feeling of having been worse off than your sibling may arise from (perceived) differential parenting (for early life adversity) [66, 67], differences in innate abilities and/or unequal parental resource investment (for education) [68], and sibling deidentification in the proband (i.e. actively seeking to differentiate themself from their sibling; for alcohol use) [69]. Siblings may use each other as a reference point, which in the case of upward social comparisons (i.e. comparisons to a perceived ‘superior’ other) may lead to experiences of unfairness and inequity [24]. Such upward social comparisons have been shown to have the most detrimental effects on depressive and anxious psychopathology by feeding into dysfunctional beliefs about the self [25]. Of note, among risk factors for which we found significant additional impact of considering proband levels, evidence was less strong for low income/education given that these associations were not significant after correcting for multiple testing.

We did not find evidence for an added effect of taking into account proband risk over-and-above or in combination with a sibling’s individual risk for any of the other psychosocial risk factors that we tested, which is in line with previous findings (age, gender) [5, 39, 40]. For instance, recent negative life events, for which we reported moderate proband-sibling resemblance in a previous study [8], were not associated with sibling psychopathology. One possible reason for that is that we measured recent life events in a sample of relatively older aged adults, in which factors beyond the family environment (such as individual, rather than familial, recent negative life events) may have a larger impact. As such cross-sibling effects of recent negative life events on affective psychopathology may be less likely. This is in line with evidence from behavioral-genetic research [6, 7] suggesting an increased role across the lifespan for individual environments and unique risk and protective factors in shaping behavioral, psychological, and personality features.

Overall, results were highly similar between the two outcomes with regard to the direction of associated risk factors, but lifetime diagnosis (dichotomous) showed fewer associated factors within siblings and no associated proband main effects or sibling × proband interaction effects for any of the risk factors, as compared to current symptomatology (continuous). This suggests that the continuous outcome may have provided deeper resolution and/or higher statistical power. This is particularly the case with regard to recent life risk factors and the dichotomous outcome, since we investigated associations between a risk factor that occurred in current/recent life (e.g. unemployment) with a disorder that may have occurred years before the time of assessment, whereas the continuous outcome referred to current symptoms. This may have reduced power to detect proband main effects and/or sibling × proband interaction effects of recent life risk factors and the dichotomous outcome by diluting effects in both siblings and probands.

Strengths of the present study include the sibling structure of the data, which has the advantage that sibling relationships contain a higher shared proportion of (early) environmental factors as compared to parent–offspring relationships; the relatively older age of the sample, which allows for the examination of siblings’ more definite clinical profiles and interindividual discrepancies between siblings that emerged across the lifespan; and the wide variety of assessed psychosocial risk factors. However, the present study is not without limitations. First, as this study only used cross-sectional data, no conclusion can be drawn with regard to the ordering of effects. In particular with regard to some of the lifestyle risk factors, such as smoking or hazardous alcohol use, the association with depressive and anxious psychopathology may be bidirectional [19, 20, 70]. Second, although this study used a relatively large clinically relevant sibling sample of 380 proband-sibling pairs, we may have had insufficient power to detect the examined sibling × proband risk factor interaction effects: based on the assumption that the interaction effect is half size of the main effects, 16 times the sample size is required to estimate an interaction than to estimate a main effect [71]. Third, retrospective self-report measures of life adversity may have been confounded by participants’ differential recall accuracy and current mood. However, we deem the impact low because previous NESDA and other studies showed that these measures had adequate temporal stability and were not critically affected by respondents’ current mood [72‐78]. Fourth, this study explored as a ‘bench-mark’ the added impact of individual psychosocial risk factors in probands, over-and-above or in combination with risk factors in siblings. Given the relatively exploratory nature of our study and the fact that several of the studied risk factors were correlated and may, therefore, explain overlapping portions of the symptom variance, we used separate analytical models for each risk factor. For future research, it would definitely be worthwhile to investigate the impact of individual risk factors in probands on top of a broad set of risk factors in siblings. Fifth, the fact that the present study was designed to include a high-risk sample of probands with a lifetime depressive and/or anxiety disorder and their siblings limits the generalizability of the findings to the general population.

To conclude, this study confirmed the association of a broad range of psychosocial risk factors with depressive and anxiety symptoms and disorders in siblings of probands with a lifetime depressive or anxiety disorder. However, importantly, we demonstrated that not only the risk factors within these at-risk individuals are important, but also those within their relatives: the individual-level risk factor in the proband, in itself or in combination with the individual-level risk factor in their sibling, had additional value for siblings’ psychopathology over only considering the individual-level risk factor in the sibling. Even though percentages of additional explained variance were small, previous research [79] has argued that small effects are the norm, rather than the exception, and form an indispensable foundation for cumulative psychological science: small effects may still have substantial direct consequences for individual mental well-being, especially for effects that accumulate over time and at scale such as childhood trauma [80, 81] and low income [82]. Our findings underscore the importance of weighing risk factors within a family context, as whether or not risk factors are shared with other siblings in the family may provide relevant information on the individual risk of depressive/anxious psychopathology. Future studies are needed to identify the exact mechanisms explaining the additive impact of weighing risk factors within all siblings in the family. Moreover, given the recent findings of little specificity in familial transmission in specific classes of psychiatric disorders [83‐85], future studies may want to examine whether our findings extend to other risk factors and/or psychiatric conditions.