We conducted a case–control observational study that included all eligible children from the Slovenian National Registry of Childhood T1D and their parents or caregivers, as well as a healthy control group of peers. The main inclusion criteria for the cases were an age between 8 and 15 years and a duration of T1D of more than 1 year. The exclusion criteria were intellectual disability and active psychosis. We invited the study group by mail, followed by a telephone invitation by their diabetologist before attending the regular three-monthly check-up. The control group consisted of children of the same age group who did not have T1D. They attended five randomly selected elementary school from across the country. The nature of the study was explained to the parents and children by one of the researchers or school counsellors. Parents were also given a written information sheet explaining the study.

Inclusion in the survey began in July 2015 and ended in December 2019. The invitation was sent to 124 families of children with T1D, 101 chose to participate. 380 children and one of their parents or caregivers were invited to the control group, 115 chose to participate, 9 of whom withdrew during the study. Participation was voluntary and anonymous, and all participants and their parents signed a consent form prior to participation. The recording of the interviews and the assessments were conducted at the university pediatric clinic (the cases) or at school before the start of classes (control group).

We performed descriptive comparisons between cases and controls in terms of age, sex, attachment classification and psychopathology, as well as comparisons between the securely and insecurely attached groups and the female and male sex in terms of psychopathology. T-tests for independent samples, Mann–Whitney U-tests, Pearson chi-square tests and Fischer exact tests were used in the IBM SPSS statistical package for continuous or categorical variables. The linear regression model in SPSSS and the statistical program R were then used to predict psychopathology (CBCL). The independent variables used in the models were: T1D, a dichotomous attachment classification (secure or insecure), presence of disorganized attachment, attachment-related anxiety of the parents (ECR-RS anxiety) and stressful life events of the child (LITE-P). In a second phase, sex, age, and interactions between the variables were added as described above.

Before multivariate modeling, we assessed bivariate associations between psychopathology (CBCL) and the following candidate predictors: T1D, attachment classification (secure vs. insecure), disorganized attachment, parental attachment-related anxiety (ECR-RS anxiety), child stressful life events (LITE-P), age, and sex. Multivariate linear regression models were used to assess the correlations between the independent variables and psychopathology. The independent variables were selected according to two criteria: (i) statistically significant relationship with the dependent variable and (ii) selection of variables based on theoretical background (enter method)—T1D, child attachment classification (secure/insecure) and disorganized attachment, parental attachment-related anxiety (ECR-RS anxiety), and child stressful life events (LITE-P). Second, the stepwise procedure was used to better understand the relationships between additional predictors influencing the presence of psychopathology. In the stepwise procedure, our model included all main effect terms and all combinations between pairwise variables as interaction terms. The stepwise procedure was described in our previous reports (Turin et al., 2021). Age was entered as a continuous variable. For significant interactions involving continuous moderators, we used simple slope analyses to estimate conditional effects at the mean of the moderator and at ±1 SD.

207 children aged 8–15 years (mean 11.74 ± 2.08) took part in the study. 101 children had T1D (49.5% female), and 106 children (55.7% female) were healthy controls. There were no statistically significant differences between the groups of children with T1D and their healthy peers in terms of demographic data and attachment security, the only difference was the educational level of both parents, which was statistically significantly higher in the control group (p < 0.01) (Table 1) (Turin et al., 2021). This comparison has been previously reported (Turin et al., 2021). The comparison of the two groups also showed no statistically significant differences in terms of psychopathology in all categories assessed by CBCL or YSR (p > 0.05) (Table 2).

Of the children who took part in the survey, 121 (59%) were securely attached. The majority of insecurely attached children were dismissing (39%) in a 3-way classification. In a 4-way classification, 16% of the children were disorganized (Table 3). Among the securely attached children, there was a higher percentage of girls (62%). The difference was statistically significant (p < 0.01). No additional significant differences were observed between securely and insecurely attached children with respect to general demographic characteristics (Table 4). Disorganization in attachment was statistically significantly associated with parental attachment insecurity (higher scores for attachment anxiety and avoidance, Table 4).

Based on the preliminary data analysis, which showed a high agreement between children’s attachment to the mother and attachment to the father (95% of children showed the same two-way classification for both parents), only attachment to the mother (N = 205) was used in further analysis to avoid multicollinearity between the independent variables. Thus, two cases were excluded because we only had information on attachment to the father. There was also a moderate association between parents’ attachment anxiety and parents’ attachment avoidance (r (91) = 0.43, p < 0.001). We only included attachment anxiety in our models for the same reason as later, to avoid multicollinearity.

When comparing psychopathology, insecurely attached children had higher scores in several categories in the CBCL/YSR, but only the CBCL/YSR category withdrawn/depressed and the CBCL category social problems reached statistical significance (p < 0.05) (Table 5). In the 4-way classification, children classified as disorganized had higher scores in all domains of the CBCL/YSR, many of which reached statistical significance: CBCL/YSR withdrawn/depressed (p < 0.05), CBCL/YSR social problems (p < 0.01), CBCL/YSR internalizing problems (p < 0.05; p < 0.01), CBCL/YSR total score (p < 0.05; p < 0.01), CBCL somatic complaints (p < 0.05), CBCL aggressive behavior (p < 0.05), YSR anxious/depressive (p < 0.01), and YSR thought problems (p < 0.01) (Table 5). The agreement between parents and adolescents (CBCL and YSR for the same scales) measured by the Pearson correlation coefficient was statistically significant in all problem scales, ranging from 0.24 to 0.43 (p < 0.01).

A comparison of psychopathology between the sexes showed that boys scored higher than girls in all CBCL domains, apart from somatic complaints. The difference was statistically significant for thinking problems (p < 0.011), attention problems (p < 0.01) and the total score for psychopathology (p < 0.05). Although girls tended to report more internalizing and boys more externalizing and attention problems, the YSR sex comparison only showed a statistically significant difference for somatic complaints, where girls had higher scores (p < 0.05) (Table 6).

There were significant correlations between the child’s psychopathology and the parents’ attachment insecurity (anxiety (r = 0.18 to 0.23) and avoidance (r = 0.14 to 0.28)). Parental attachment anxiety was statistically significantly related to the anxious/depressed scale and internalizing problems on both the CBCL (r = 0.21–0.22; p < 0.01) and YSR (r = 0.18–0.21; p < 0.05) scales, social problems and the total psychopathology score on the CBCL scale (r = 0.20–0.21; p < 0.01). Parental attachment avoidance was statistically significantly related to anxious/depressed, withdrawn/depressed, social problems, internalizing problems, and total score on both the CBCL and YSR scales (r = 0.16–0.304; p < 0.01 or p < 0.05) and with somatic complaints, thinking problems and rule-breaking behavior on the YSR scale (r = 0.24–0.28; p < 0.01 or p < 0.05) (Table 7).

Parents’ reports of traumatic life events were statistically significantly related to psychopathology as rated by parents, but not by children. Children’s ratings of stressful life events, however, were associated with some psychopathology scales (total psychopathology score, as well as anxious/depressed, thought problems, rule-breaking behavior, and the internalizing scale) (Table 8).

Due to the higher number of complete cases and to avoid multicollinearity between parent- and child-reported measures, only parent-reported psychopathology (CBCL) and adverse life events (LITE-P) were included in the subsequent analyses. The results of the multivariate linear regression models showed that the significant predictors of psychopathology were attachment insecurity, especially attachment disorganization, higher parental attachment anxiety and traumatic life events. T1D was not a statistically significant predictor in this model (Table 9).

In the subsequent analyses, we constructed multivariate linear regression models including sex, age, and selected interaction terms, guided by a biopsychosocial framework in which psychopathology is understood to arise from interactions between biological and psychosocial factors across development (Cicchetti and Rogosch, 2002).

The results showed that the child’s male sex and the presence of T1D were significant predictors of psychopathology. Two interactions between independent variables showed statistical significance, namely the interaction between parental attachment anxiety and age and between parental attachment anxiety and traumatic life events (Table 10). Simple slope analyses were calculated for these interactions, which are shown in Figures 1, 2.

The analysis of interactions showed significant correlations between parental attachment anxiety and psychopathology as a function of age. Middle-aged and older children whose mothers showed more attachment anxiety had higher psychopathology scores. When parents reported high levels of attachment anxiety, there was also a significant relationship between age and psychopathology (Figure 1).

The interaction analysis also showed that children with an average and higher number of traumatic life events whose mothers had higher attachment anxiety had higher psychopathology scores. When parents had average or higher levels of attachment anxiety, there was also a significant relationship between the number of traumatic life events experienced by the child and the child’s psychopathology (Figure 2).

In our cohort, 59% of the children were classified as securely attached. Previous studies using the same measures have reported a slightly higher proportion of securely attached children (66%) (Costa-Cordella et al., 2020). Among the insecurely attached children, the majority were dismissive (39%), which is also slightly higher than in the study by Shmueli-Goetz et al. (2008) (30%), 2% were classified as preoccupied, which is lower than in the previously mentioned study (4%). In our sample, 16% of children were disorganized in the 4-way classification, which is much higher compared to the 4% in the study by Shmueli-Goetz et al. (2008).

Our results demonstrate a significant association between disorganized attachment and multiple domains of psychopathology in both groups, given that no differences in attachment representations were observed between children with T1D and healthy controls (Table 1). Previous studies have identified insecure attachment as a risk factor for internalizing and externalizing psychopathology in childhood and adolescence (Fearon et al., 2010; Groh and Narayan, 2019). Within insecure attachment, a disorganized attachment style between child and parent has been associated with neurological vulnerability in the child and thus with a range of later social and cognitive difficulties as well as psychopathology (van Ijzendoorn et al., 1999). Disorganized attachment is related to an incoherent internal representation of the same parent who simultaneously evokes the need for closeness and anxiety (Green and Goldwyn, 2002). This results in difficulties in stress and emotion regulation, impaired mentalization (the ability to understand mental states of self and others), and interpersonal dysfunction, which together increase the risk of psychopathology (Luyten et al., 2020).

Our findings suggest that, when controlling for other variables, child attachment security may play a lesser role than factors such as parental attachment security, traumatic life events, age, sex, and the presence of a chronic illness like T1D. Parental attachment insecurity appears to be a more important risk factor than child insecurity, which may indicate that the parents’ emotional well-being and their ability to regulate the child’s stress are more important than the intrinsic abilities of the child itself (Nelson et al., 2009).

However, the population included in our study was non-clinical in terms of psychopathology. It is possible that in a clinical psychiatric sample of adolescents, the influence of attachment on psychopathology would be stronger. This can be inferred from a study of psychiatric inpatients that included adolescents with psychiatric disorders such as depression, obsessive-compulsive disorder, and oppositional defiant disorder which showed an inverse distribution of attachment security (30.4% secure and 69.5% insecure attachment) (Venta et al., 2014).

There were no differences in psychopathology when comparing children with and without T1D. However, findings in the literature remain inconsistent. Some studies have reported a higher incidence of mental disorders, such as depression, anxiety and eating disorders, among children with T1D, while others have found a lower or similar incidence compared with healthy children (Sivertsen et al., 2014; Dybdal et al., 2018). These discrepancies may be partly explained by differences in age range, methods used to assess psychopathology, and variations in access to specialized psychosocial care.

Children and adolescents at different developmental stages face distinct challenges in both mental health and diabetes management (Cicchetti and Rogosch, 2002; Berg et al., 2017). Younger children typically depend on their parents for daily diabetes care and may show fewer overt signs of depression or anxiety (Berg et al., 2017). In contrast, adolescents enter a critical period of biological, emotional, and social change, during which they increasingly assume responsibility for self-management (Bombaci et al., 2024). Puberty, hormonal fluctuations, and psychosocial stressors can impair metabolic control and increase disease-related stress, making this a particularly vulnerable period for psychological difficulties (Bombaci et al., 2024). As further discussed, our results highlight the importance of age and sex differences in the relationship between T1D and psychopathology.

Some studies focused on subclinical symptoms, while others assessed formal psychiatric diagnoses, which may have further contributed to divergent results (Dybdal et al., 2018; Telo et al., 2018). One large population-based sample showed that mental health symptoms were more common in adolescents with T1D than in healthy controls, whereas diagnoses of common mental disorders were not more prevalent, suggesting increased subclinical psychological distress rather than overt psychopathology (Telo et al., 2018). A Slovenian study involving 126 adolescents with T1D and 499 healthy adolescents showed that the prevalence of suicidal thoughts, suicide attempts and self-harming behavior was lower among the adolescents with T1D than among healthy peers (Drobnič Radobuljac et al., 2009). At the same time, the prevalence of eating disorders was higher among female adolescents with T1D (Drobnič Radobuljac et al., 2013). In this cohort, most adolescents were treated with insulin pumps, reflecting well-organized, multidisciplinary diabetes care, which may contribute to favourable mental health outcomes (Turin et al., 2021). Nevertheless, within the same T1D cohort, suicidal behavior was associated with poorer metabolic control (Drobnič Radobuljac et al., 2009), indicating that vulnerable subgroups remain at increased risk. Consistently, larger population-based studies have demonstrated an association between poor metabolic outcomes and psychiatric comorbidities (Sildorf et al., 2018).

Our results showed that the presence of T1D has no significant effect on psychopathology when the individual variables were analyzed. In contrast, when interactions among variables such as parental attachment anxiety, child age, and traumatic life events were considered, T1D was associated with an increased risk of psychopathology. Male sex and T1D were associated with higher CBCL total scores (main effects). Consistent with the earlier discussion on developmental stage, age moderated the association between general anxiety and CBCL scores: the relationship was significant at the mean age and +1 SD, but not at −1 SD, highlighting adolescence as a period of increased vulnerability to psychological difficulties in the context of T1D (Bombaci et al., 2024).

The association between maternal attachment anxiety was also statistically significant in children with average or higher frequency of negative life experiences reported by parents. According to the results, special attention should be paid to boys with T1D who have been exposed to more negative life events and whose mothers appear to be less securely attached. Since the association between attachment anxiety and psychopathology was only observed in middle-aged (Mean age 11.7 years) or older children, special attention should be paid to adolescent boys compared to younger children.

A previous study on the same cohort of children also highlighted the sex differences in metabolic outcomes of T1D. The results showed that in boys there was a positive correlation between parental attachment style and glycemic outcomes. The more insecure the parental attachment, the higher the mean Hb1Ac, HbA1c variability and the lower the TIR (time in required glycemic range). The reverse relationship was observed in girls (Turin and Drobnič, 2021). This may be explained by the transfer of responsibility for disease management from parents to adolescents, with more insecurely attached parents possibly being more anxious and controlling, which may be beneficial in younger children but may hinder autonomous motivation and information sharing in adolescents (Herzer and Hood, 2010). A trusting relationship has been shown to be a protective factor for good glycemic control in adolescence (Berg et al., 2017). In securely attached girls, the transfer of responsibility for T1D management may have occurred earlier, possibly exceeding their maturity and leading to suboptimal metabolic outcomes (Klemencic et al., 2023).

A strength of the study is the recruitment of nearly the entire cohort of Slovenian children with T1D aged 8–15 years, using a validated qualitative interview analyzed by three independent coders to assess attachment. This is the first Slovenian survey that allows comparisons of attachment profiles with international studies.

A limitation of the study is a rather homogeneous nature of the cohort, as all children identified as white and most used insulin pumps, limiting generalizability. We used a qualitative approach to assess children’s attachment to their parents, but parents’ attachment was only assessed using a self-report questionnaire, which, despite its reliability and validity, is at risk of over- or under-reporting. Our cohort was not clinical in terms of psychopathology, and we only had information on symptoms based on self/parent evaluations. Our results also showed an association between mothers’ attachment anxiety and parent-rated psychopathology, suggesting that lower maternal stress regulation influences children’s psychopathology, but may also indicate that mothers with higher attachment anxiety exaggerate their children’s symptoms. The exclusive use of parent ratings was aimed at increasing the power of the study by obtaining more CBCL reports.