In this study, adolescents and their best friend from a relatively well-functioning Dutch sample were followed longitudinally across two years (three waves). The current study focused on their development of support and depressive symptoms, with the goal to unravel individual from dyadic level effects of the association between best friend support and subsequent depressive symptom development. Results showed no associations at the individual level. However, dyadic level associations of support with subsequent depressive symptom development were found, and this association depended upon the initial levels of depressive symptoms experienced by the dyad. That is, for dyads with few depressive symptoms, higher levels of dyadic support predicted that both members of the dyad would show relative decreases in depressive symptoms over time when compared to dyads with lower levels of support. The opposite was found for dyads characterized by more depressive symptoms. That is, for these dyads higher levels of support predicted relative increases in depressive symptoms over time for both members of the dyad, compared to dyads with lower levels of support.
Our findings on the absence of a main effect of support on adolescent depressive symptoms concur with previous studies that found no associations between friendship support and depression symptom development (Bámaca-Colbert et al., 2017; Ling et al., 2022; Lyell et al., 2020). However, several other studies did find associations between close friend support and depressive symptoms (Finan et al., 2018; Rueger et al., 2016). Because these previous studies did not distinguish individual from dyadic level effects, it is not possible to determine at which level effects actually occur. Therefore, it is possible that effects found in other studies are attributed to individual processes when in reality dyadic processes might be involved.
Furthermore, our finding that the association of support with subsequent depressive symptoms depended upon the initial levels of depressive symptoms and that this effect was only found on the dyadic level (and not on the individual level), moves beyond previous studies in two ways. First, it shows the necessity to consider dyadic level effects when studying the association between friendship support and subsequent depressive symptoms. In fact, in our study, no associations of support and subsequent depressive symptoms were found at the individual level. Although previous studies did not differentiate between dyadic and individual effects, some studies have suggested that dyadic processes such as depression contagion via peer influence or selection effects might play a role in the development of depressive symptoms within adolescent friendships (Boersma-van Dam et al., 2019; Brechwald & Prinstein, 2011; Giletta et al., 2011; Van Zalk et al., 2010). The present study confirmed this suggestion and emphasizes that part of the support-depression association may be due to either selection or within dyad processes, or both, that may make friendship dyads similar in the appreciation of support in their experienced mental state (Mankowski & Wyer, 1996; Van Zalk et al., 2010).
The second way our results moved beyond the current literature regards the nature of the dyadic level association. Although results need to be replicated first before drawing firm conclusions, our finding that the effect of best friend support on subsequent dyadic depressive symptom development was different for dyads with more depressive symptoms versus dyads with few initial dyadic depressive symptoms (i.e., initial dyadic depression levels moderate the association), is in line with previous theoretical suggestions (Gariepy et al., 2016; Pfeiffer et al., 2012). For dyads with low initial levels of depressive symptoms, dyadic support seems to have a protective effect on depressive symptom development of the dyad. This finding is supported by previous theoretical considerations and empirical evidence from two meta-analyses (Chu et al., 2010; Cohen & Wills, 1985; Rueger et al., 2016). Both meta-analyses found weak associations between higher levels of friend support and higher levels of well-being or lower levels of depressive symptoms; our study implies that these effects may be dyadic and possibly dependent on initial depression levels. Friendship support might ease emotional distress and improve mental well-being, but only when symptoms are mild.
The opposite was found for dyads with relatively more depressive symptoms. For these dyads, support was associated with a relative increase in depressive symptoms over time. Recent studies have aimed to explain this seemingly counterintuitive finding by the process of ‘co-rumination’. Co-rumination is a maladaptive communication style between friends, used to regulate their emotions. When friends co-ruminate, they discuss a problem they are dealing with thereby rehashing the problem and excessively focus on their negative feelings and emotions instead of possible solutions (Rose, 2002; Rose et al., 2007). At first, co-ruminating tends to make friends closer, as it is associated with perceived increased friendship quality and increased support (Ames-Sikora et al., 2017; Rose et al., 2007). However, co-rumination might also exacerbate adolescents’ depression experiences. Indeed, the mutual tendency to dwell on problems within friendship dyads has been associated with (increasing) depressive symptoms (Rose et al., 2007, 2014). Although never tested before (to our knowledge), one might speculate that co-rumination may manifest particularly in friendships characterized by higher depression levels. Similar to rumination, where individuals suffering from depression dwell on their problems internally (Nolen-Hoeksema, 1991), co-rumination might be characteristic of friends who show more depressive symptoms and less so of friends with relatively few depressive symptoms. As co-rumination goes hand in hand with friend support (Ames-Sikora et al., 2017; Hankin et al., 2010; Rose et al., 2007, 2014), this might explain why support from friends in certain instances might exacerbate depressive symptoms.
Thus, the current study showed—in a sample with generally mild depressive symptoms—that in order to understand processes that may explain why friendship support associates with depressive symptoms, the dyad as a whole should be considered as well as initial depression levels experienced by friends. It is important to note that all results were controlled for gender as previous literature showed mixed support for gender differences in the levels of and associations between friend support and experienced depressive symptoms (Chu et al., 2010; Gregory et al., 2020; Rose & Asher, 2017; Rose & Rudolph, 2006; Rueger et al., 2010, 2016).
Although not of main interest to our study and not officially tested, we also found that friends seem to become more alike in their level of depressive symptoms over time (as indicated by increasing ICCs). This finding is in concurrence with other studies (Boersma-van Dam et al., 2019; Schwartz-Mette & Smith, 2018; Van Zalk et al., 2010) and emphasizes that early monitoring of friends with elevated depression symptoms is warranted.
Limitations
Two important limitations of the current study should be noted. First, the longitudinal nature of the data collection with the complex full-family design and the inclusion of only stable, same-gender friendships across the studied period may have resulted in a rather homogenous sample that may not generalize to the broader Dutch population. For instance, the sample included an over-representation of adolescents from families with a relatively high socioeconomic status and included very few adolescents from other ethnic backgrounds than Dutch. Although two recent studies that compared levels of depressive symptoms between a) native Dutch and minority youth and b) children living in the Netherlands from lower and higher socioeconomic backgrounds found no differences in depressive symptom levels for these youth in early adolescence (Buil et al., 2022; Horoz et al., 2022), some studies in other countries did find such differences (Brown et al., 2007; Wickrama et al., 2009). Moreover, it is expected that approximately 4% of adolescents in a western society have clinically elevated levels of depressive symptoms, which is 50% higher as in our sample (2.7%) (Costello et al., 2006; Keyes et al., 2019). Therefore, caution is warranted when generalizing findings.
Second, both best friend support and depressive symptoms were measured via self-reports. Although one might argue that subjectively perceived support and depressive symptoms are (most) important when studying the association between friendship support and subsequent depressive symptoms, the experienced level of depressive symptoms might influence the perception of friend support. Therefore, future studies are encouraged to replicate our findings including more objective measures of support and depressive symptoms. Lastly, we used a correlational design in our longitudinal study, and our findings by no means imply causal effects of support on depressive symptom development.
Implications and Future Research
To understand the association between friendship support and subsequent adolescent depressive symptoms it is recommended that future research takes on a multilevel approach and searches for associations at all levels. Also, future research could be directed at understanding why support and depressive symptoms are associated on a dyadic level and why friendship dyads seem to become more alike in their depressive symptoms. It could consider processes of selection and within dyad processes that may underlie the association of friendship support and subsequent dyadic depressive symptoms over time. For instance, it might be that some dyads have a shared higher susceptibility to peer influence that causes similarity in their depressive symptoms.
Furthermore, it is recommended that researchers include gender as a variable of interest and test for potential differential effects for different genders. Although previous research is somewhat mixed on the effect of gender on the association between friend support and depressive symptoms, studies consistently show that girls are twice as likely to develop depressive symptoms than boys (Salk et al., 2017). Recent research considers co-rumination to be one of the mechanisms associated with this heightened risk (Rose, 2002; Stone et al., 2011). Co-rumination seems to be associated with more depressive symptoms regardless of gender (Rose, 2002; Spendelow et al., 2017), however several studies report girls to engage more in co-rumination than boys, resulting in higher rates of depression for girls (Felton et al., 2019; Rose, 2002; Rose et al., 2007; Spendelow et al., 2017). Thus support within girl-dyads might be characterized more by co-rumination than support within boys-dyads and therefore associated with more depressive symptoms for girls.
Moreover, studies might want to investigate the effects of support in samples of adolescents with clinical depressive disorders. Interpersonal theories of depression and empirical research suggest that adolescents with elevated depressive symptoms are less likely to form stable friendships and experience a decrease in support from friends over time (Coyne, 1976; Ren et al., 2018; Rudolph et al., 2008). Because adolescents suffering from clinical depression already experience more difficulty getting support from friends than their peers with few depression symptoms, it would be interesting to examine how the friend support they do experience is related to their depressive symptoms from a dyadic perspective. It could be that adolescents with more severe depression symptoms experience additional difficulties because in the friendships they (still) have the support is related to even higher depressive symptoms.
Lastly, future research should focus how the results generalize to different sources of social support as each of the different sources of support might have their own specific contribution to adolescent depression development. For example, previous studies found that adolescents who perceive (increasing) parental support (Gariepy et al., 2016; Stice et al., 2004), sibling support (Finan et al., 2018) or teacher support (Reddy et al., 2003) showed decreases in depressive symptoms. Also the unique contribution of each source of support should be taken into account as support is almost always given in the context of other sources of support and possible compensatory effects might affect the results (Gregory et al., 2020; Milevsky & Levitt, 2005).
If the findings are replicated in follow-up studies, the following practical implications can be considered. It may be an option for preventive and monitoring measures to incorporate dyadic features and differentiate between dyads experiencing relatively many or few depressive symptoms. Seeking support from peers is often encouraged in universal prevention programs or incorporated in interventions targeting adolescents already suffering from depressive symptoms (Ali et al., 2015; Jones et al., 2018). National and international campaigns, such as ‘hey het is ok’ – initiated by the Dutch government in 2021—or the WHO’s ‘Depression: let’s talk’ campaign—that stimulate interpersonal communication about depression, encourage people to seek support from friends, and encourage friends to listen to and assist their friend with elevated depression symptoms, might benefit from incorporating information about how friends can provide healthy support (i.e., advise that goes beyond ‘listening’). Without nuance, such campaigns might be counter-effective and actually might do more harm than good for some friends. This is of particular importance because adolescents themselves state to prefer informal help from friends and experience barriers when seeking professional help (Singh et al., 2019). Moreover, if follow-up studies indeed show that these processes occur systematically at a dyadic level, conventional treatment programs for depressive symptoms, such as cognitive behavioral therapy (CBT; David-Ferdon & Kaslow, 2008) and interpersonal psychotherapy (IPT; David-Ferdon & Kaslow, 2008), might also benefit from adding dyadic aspects to their therapies. For example, these programs could provide psychoeducation on how to seek and provide healthy support and on how to interact with friends in a constructive and helpful way. It would be worthwhile to consider incorporation of friends into a CBT or IPT program and provide therapy on the level of friendship dyad or provide group therapy that includes friendship dyads.
In sum, our results indicate that for adolescent friends who both experience relatively few depressive symptoms, promoting supportive dyadic relations could be beneficial. However, adolescents who form a friendship dyad characterized by relatively more depressive symptoms compared to the other adolescents in our sample, support from a best friend might exacerbate symptoms. If replicated, this potential contagion effect of the friendship might be considered in the prevention and treatment of an adolescent with depressive problems.