Perceived carer skills and psychological distress in fathers and mothers of adults with anorexia nervosa: a longitudinal study Authors: Melahat Nur Akkese, Johanna Keeler, John Hodsoll, Suman Ambwani, Hubertus Himmerich, Katie Rowlands, Valentina Cardi, Janet Treasure Published: 2025 ABSTRACT BackgroundParents often remain the primary carers for their adult children with anorexia nervosa (AN) for an extended time. Carer strain and unhelpful behaviours have been extensively investigated as contributors to the maintenance of the illness and suggested several differences between fathers and mothers. However, the skilful dimensions of parental caregiving have been overlooked, and no longitudinal research has examined the caregiver abilities and psychological well-being of parents during their child’s transition from an intensive treatment. This exploratory study compares the trajectories of change in perceived carer skills and psychological distress between fathers and mothers of adult patients with AN over an 18-month period during the transition from intensive hospital care.MethodsFifty two fathers and 224 mothers of adult AN patients were included in this study. A secondary analysis was conducted on data from the TRIANGLE trial using linear mixed-effects models. Outcomes were the total and subscale scores from the Caregiver Skills Scale (CASK) and the Depression, Anxiety, and Stress scale (DASS-21) in fathers and mothers at baseline, 6, 12, and 18 months.ResultsPost-hoc tests revealed that fathers and mothers did not show significant differences in the CASK scores at any time-point (except for higher paternal “Self-Care” and “Biting-your-tongue” skills at baseline). However, the trajectories of change in their abilities differed between parents. Mothers showed significant increases over time in the overall CASK and in several of the subscales from baseline to month 18 (but not months 6 or 12), whereas fathers did not show significant changes in the total or subscale scores at any time point. In contrast, the changes in distress levels (DASS-21) followed similar trajectories for fathers and mothers over time.ConclusionsFathers and mothers may have distinct needs in order to enhance their caregiving skills in the long-term. Although the findings require validation in future research involving father-mother dyads, effort should be made to ensure that carer support is tailored to be fit-for-purpose for both fathers and mothers.Trial registrationThe TRIANGLE trial whose parental data was used in this study was pre-registered on the ISRCTN registry (ISRCTN14644379).Supplementary InformationThe online version contains supplementary material available at 10.1186/s40337-025-01509-2. Background Parents are usually the main carers involved in supporting individuals with AN, both during the early stages of the illness and throughout the lifespan in those with a longstanding course of AN [1, 2]. A carer can be defined as a person who provides informal, unpaid care and support to a family member or loved one with a physical and/or mental health condition (The Princess Royal Trust; www.carers.org) In the context of eating disorders (EDs), the caring role can include (i) medical (e.g., seeking medical help), (ii) practical (e.g., providing meal support), (iii) emotional (e.g., communicating with the patient about their difficulties), (iv) logistical (e.g., accompanying the patient to hospital for treatment), or (v) financial (e.g., funding specialist care) support to the individual with an ED [3, 4]. Thus, regardless of the offspring’s age (i.e., adolescent or adult), parents report a significant caregiver strain related to their child’s severe clinical state, their beliefs about the negative consequences of AN on their child, increased time spent on caring tasks, and higher levels of anxiety and depression [3, 5, 6]. These factors form part of the cognitive interpersonal maintenance model of AN [7]. According to the interpersonal component of this model, carers’ own vulnerabilities and traits (e.g., stress sensitivity, obsessive-compulsive personality traits) may predispose them to challenges in coping with the demands of their caregiving role and lead to emotional and behavioural reactions to AN (i.e., high levels of expressed emotions, increased criticism, over-protection, and accommodating and enabling behaviours). These not only contribute to the maintenance of AN symptoms but also perpetuate parental distress by creating a vicious circle [7, 8]. Therefore, skills-based interventions (e.g., Experienced Caregivers Helping Others – ECHO, the New Maudsley Collaborative Care) have been developed to empower carers with helpful behaviours and skills to maximise their effectiveness in their supporter role by providing them with information about EDs and training in caregiving skills (e.g., self-help materials, workshops, coaching sessions with professionals, carers, or people with lived experience of EDs) [9–12]. Several studies have demonstrated the effectiveness of these interventions for parents and their adolescent or adult loved ones [9, 11, 13]. The transition from intensive hospital care to home typically presents challenges for parents as well as patients. The recent qualitative studies [14, 15], for example, revealed that most parents experience mental and/or physical health problems and feelings of isolation, guilt, anxiety, or exhaustion related to their loved ones’ deteriorated clinical state, perceived AN relapses, uncertainty over the future of their loved ones, caregiving demands (e.g., giving up work, having to juggle house and work responsibilities, time constraints), and inadequate aftercare support by health services. They also conveyed a need for more tailored and needs-based support, including guidance about post-discharge, skills trainings, and attention to their own well-being. Thus, aftercare interventions were developed to prevent relapse and to facilitate the transition process for both carers and patients, such as (online) self-help interventions (e.g., ECHO, an Internet-based adaptation of the Maudsley Model of AN for Adults—iMANTRA) [16, 17]. The TRIANGLE study was a pragmatic, two-arm, multicentre randomised controlled trial that tested the efficacy of a digital guided self-management intervention for patient–carer dyads to support transition from intensive care (inpatient or day) to community [18–20]. In this study, we examined whether the addition of an online hybrid (both carer and patient support) skills-sharing intervention (Experienced Carers Helping Others [ECHO]—Maudsley Model of Anorexia Nervosa Treatment for Adults [RecoveryMANTRA]; ECHOMANTRA) to treatment-as-usual (TAU) improved dyad outcomes over 18 months during the transition, compared with those receiving TAU alone. Carers were chosen by the patient, and only one carer (i.e., any family member or friend willing to provide some aftercare support) participated in the study as the primary caregiver [19]. Further details about the TRIANGLE trial are available in Box 1 (see Additional file 1) and the published primary results [19, 20]. In the primary trial, no differences were found in carer outcomes (i.e., perceived carer skills and psychological distress) between the trial arms [19, 20]. We initially hypothesized that carers in the ECHOMANTRA plus TAU arm would have improved carer skills and lower psychological distress. However, we did not specifically investigate these outcomes in the carer subsamples (e.g., fathers, mothers). Previous research has suggested differences in the caregiving experiences of fathers and mothers of individuals with AN. Mothers usually showed higher levels of psychological distress, self- and interpersonal-related strains (e.g., loss of self, relational deprivation), and more stable attendance to the treatment of their child [3, 5, 21–25]. Fathers, on the other hand, reported higher scores of alexithymia, less time spent for daily caring tasks, lower levels of eating disorder-related (ED) strain, and less engagement with treatment, but greater self-care and the tolerance and management of unpleasant emotions [3, 5, 21, 23, 25]. Furthermore, qualitative studies have reported that mothers tended to exhibit accommodating behaviours (e.g., adapting to their child’s food-related rituals) and emotional coping responses to their child’s illness (e.g., emotional overinvolvement). Conversely, fathers tended to use cognitive and avoidant coping strategies (e.g., focusing on work to distract themselves from the situation, leaving most of the illness management to the mother), to rely on mothers as the experts in understanding the child during the recovery, and to be more preoccupied with practical support tasks (e.g., food shopping) [4, 26–29]. Nevertheless, there remains a limited understanding of whether the skilful caregiving experiences and psychological well-being of parents show similarities and/or differences during the transition of their adult child from intensive care. Therefore, this secondary exploratory analysis of the data from the TRIANGLE study examined whether fathers and mothers of adults with AN showed different patterns of change in perceived caregiving skills and psychological distress during the transition process. We compared (i) the trajectories of change in perceived caregiver skills (CASK) and psychological distress (DASS-21) between fathers and mothers nominated by patients as their primary caregivers over 18 months during the transition from intensive treatment (inpatient or day care) and explored (ii) whether any between-group (fathers versus mothers) differences in these carer outcomes were evident at baseline, months 6, 12, and 18. Methods Study design and participants This exploratory mixed-design (i.e., within-subjects and between-groups) secondary analysis [30] examined the data from the TRIANGLE trial (Trial registration: Current Controlled Trials ISRCTN14644379) [19]. A total of 371 patient-carer dyads were recruited from thirty-one specialised inpatient or day patient ED services across the UK. Each patient-carer dyad was randomly allocated to the TAU alone or augmented with the ECHOMANTRA intervention. However, one dyad that had been mistakenly randomised was withdrawn from the study due to incomplete baseline assessments and lack of consent for trial participation, resulting in a final sample of 370 dyads. The inclusion criteria for patients and carers in the primary study were: (i) being aged 16 years or above, (ii) having DSM-5 diagnosis of AN (the Diagnostic and Statistical Manual of Mental Illnesses; DSM-5) with a BMI ≤ 18.5 kg/m2 or atypical AN on admission to an inpatient/day-care unit, (iii) having a carer who was willing to participate, (iv) consenting any time after admission to hospital until 4 weeks post-discharge, (v) having an electronic device and the internet to access the digital programme. The exclusion criteria were as follows: (i) not receiving inpatient care or not attending day care for a minimum of 3 days/week when consented to the study, (ii) having an insufficient knowledge of English, (iii) having a severe mental or chronic illness, (iv) being pregnant, (v) previously receiving treatments involving ECHOMANTRA materials, (vi) incomplete baseline measures of the dyad [20]. The eligibility criteria for inclusion in this secondary analysis were: (i) being a father or mother as the primary carer chosen by the patient, (ii) there was a reasonable age difference between the parent and the patient, to identify potential data entry errors. Of 370 patient–carer dyads in the larger trial [19], 281 (∼76%) were parent–child dyads (where the carer was a parent), while 89 (∼24%) involved other types of relationships with the patient (i.e., partner/spouse, sibling, child, friend, and others). Five of the parent–child dyads were excluded because the small age difference between the parent and the child was deemed indicative of data entry errors, resulting in 276 (∼75%) patient-parent dyads (N = 52 fathers; N = 224 mothers) being included in the present study. A consort diagram describing the flow of the participation in the current study is presented in Fig. 1. Fig. 1A consort diagram describing the flow of the participation in the current study. The first part of the diagram was adapted from the primary TRIANGLE trial paper [19] to ensure consistency with the current secondary analysis A consort diagram describing the flow of the participation in the current study. The first part of the diagram was adapted from the primary TRIANGLE trial paper [19] to ensure consistency with the current secondary analysis Measures Parent measures Demographic data. Parents provided demographic data (i.e., age, sex, ethnicity, highest level of education completed, marital and employment status) in a bespoke questionnaire for carers at baseline. Caregiver Skills Scale (CASK) [31]. This tool is a 27-item self-report measure of caregiving attitudes and behaviours for coping with ED symptoms. Each item describing situations commonly related to EDs (e.g., getting drawn into arguments about the ED with the patient, urge to enquire about or check on the patient’s behaviours when feeling worried, relying solely on weight as a marker of the patient’s progress) is scored by individuals who support a person with an ED on a visual analogue scale (i.e., 0 = almost never, 100 = almost always) based on how confident they felt about the situation over the past week. The measure has six subscales: (i) Bigger Picture comprises seven items assessing the ability to adopt a positive attitude toward changes in the person’s eating disorder journey from a long-term perspective, (ii) Self-Care includes four items about taking time for self and other family members, (iii) Biting-your-tongue evaluates the capability to control the urge to enquire and prevent repetitive nagging conversations with the person with an ED through three items, (iv) Insight & Acceptance consists of three items to assess the skill to accept and manage unpleasant emotions, (v) Emotional Intelligence includes five items about the capability to discuss and manage own feelings, and (vi) Frustration Tolerance encompasses five items evaluating the skill to resolve conflict in a firm, calm, and understanding way. The total score is the sum of all items (i.e., max 2700), whereas the subscale scores are derived by calculating the average of the items in each subscale. Higher scores indicate higher levels of carer skills. The CASK has a good factorial and convergent validity and internal consistency. Both the CASK total and subscale scores were used to evaluate the caregiving skills of parents at baseline, 6-months, 12-months, and 18-months in this study. Depression, Anxiety, and Stress Scale-Short form (DASS-21) [32]. This measure evaluates depression, anxiety, and stress symptoms over the past week, through 21 items on a 4-point Likert scale (0 = did not apply to me at all, 3 = applied to me very much or most of the time). It consists of three subscales: (i) Depression (7-item) evaluates the common features of the depressive emotional state, such as hopelessness, anhedonia, lack of interest/involvement, (ii) Anxiety (7-item) assesses the physical and emotional anxiety experiences of a person, such as autonomic arousal, subjective anxiety feelings, and (iii) Stress (7-item) examines the sensitivity to chronic non-specific arousal experiences, such as being easily upset/agitated, difficulty relaxing. The total score and subscale scores are obtained by multiplying the sum of all items and the relevant subscale items by 2. The total score is used to assess overall emotional distress, with higher scores indicating greater emotional difficulties. Lovibond and Lovibond (1995) proposed the following cut-off scores corresponding to conventional severity levels (normal, mild, moderate, severe, extremely severe) for each subscale: (i) Depression: Normal = 0–9, Mild = 10–13, Moderate = 14–20, Severe = 21–27, Extremely Severe = 28 or above, (ii) Anxiety: Normal = 0–7, Mild = 8–9, Moderate = 10–14, Severe = 15–19, Extremely Severe = 20 or above, (iii) Stress: Normal = 0–14, Mild = 15–18, Moderate = 19–25, Severe = 26–33, Extremely Severe = 34 or above. The psychometric properties of the self-report scale have been confirmed in both clinical and non-clinical samples [32, 33]. Both the total and subscale scores were used to assess parent’s distress at baseline, 6-months, 12-months, and 18-months in the current study. Patient measures Demographic and clinical data. Patients provided demographic (such as age, sex) and clinical information (such as AN treatment type, clinician- and/or patient-reported weight and height at admission, years with AN, history of treatment under the Mental Health Act (MHA), number of treatments under the MHA) as part of the baseline assessment of the primary research [19, 20]. Data was mainly collected through a self-report questionnaire, but clinicians were contacted by the research team to supplement missing information where appropriate. BMI calculations (i.e., weight [kg]/height [m2]) were derived from weight and height data provided by patients themselves and clinicians at participating centres. DASS-21 [32]. As described above. The DASS-21 total score at baseline was used to assess patient distress at baseline in the present study. Eating Disorder Examination Questionnaire (EDE-Q) [34]. This 36-item self-report tool assesses the specific psychopathology of eating disorders. 14 items (items 8 and 16–28) are related to key behavioural features of EDs (e.g., binge eating, self-induced vomiting, laxative and diuretic misuse), and do not belong to any of the subscales. These items are assessed individually as indicators of the presence and frequency of the key behaviours. The remaining 22 items are used to obtain subscale and total ED scores. The four subscale scores (i.e., Restraint, Eating Concern, Weight Concern, and Shape Concern) are derived by calculating the means of the items in each subscale. The total score (i.e., Global EDE-Q) is the mean of the four subscale scores. All the resultant scores range between 0 and 6. Higher scores indicates greater frequency of ED symptoms. The EDE-Q has demonstrated good reliability [35, 36]. In the current study, the patient’s EDE-Q Global score at baseline was used as a marker of AN symptom severity. Data analysis Statistical analyses were conducted using the Statistical Package for the Social Sciences (IBM SPSS) 29.0.2 and Jamovi 2.6.44 (i.e., GAMLj3 package [Version 3.5.1]). Parametric and non-parametric tests were used to compare the baseline characteristics of the parent groups. These tests were also applied to investigate whether there were significant differences in the patient characteristics and ED severity at baseline between those who had a father or mother as their primary carer, and whether there were significant differences in demographic and clinical characteristics between the inpatient and day patient groups at baseline. When the assumptions of parametric tests were violated (i.e., non-normal distribution of continuous variables, more than 20% of cells with expected counts below 5 and a minimum expected frequency of 1 for categorical variables), non-parametric tests were conducted. Effect sizes were calculated according to the type of outcome variable and the statistical test performed. For continuous outcome variables, Cohen’s d and Pearson’s r were reported for parametric and non-parametric tests. For categorical outcome variables, Phi (ϕ) coefficient, Cramer’s V, and Kendall’s τ were reported depending on whether the variable was binary, nominal, or ordinal [37]. Further details are provided in Table 1 and Additional files 2, 3, and 4. Linear mixed-effects model (LMM) analyses were used for the main analyses to model the repeated measures whilst handling the missing data (see Additional file 5) [38, 39]. Parameters were estimated using the restricted maximum likelihood (REML) method [40]. ‘Participant’ was included as a random effect into all models, whereas other variables (i.e., parental sex and timepoint) were included as fixed effects. The DASS-21 was implemented at baseline and months 3, 6, 9, 12, and 18, whereas the CASK was implemented at baseline and months 6, 12, and 18 in the primary research [19]. For consistency, we modelled both the CASK and DASS-21 outcomes at baseline, months 6, 12, and 18. For the model equations, see Table 2 and Additional files 6 and 7. Table 1Parent baseline demographic characteristicsParent variables N Overall(N = 276)Fathers(N = 52)Mothers(N = 224)Test statisticp-valueEffect size Age 272Median (IQR)54.0 (9)55.0 (14)53.0 (7)4720.5a 0.070 r= − 0.11Ethnicity (N, %)273White264 (96.7)50 (96.2)214 (96.8)NAb 0.682 ϕ= − 0.02Others (Asian, Black, Mixed, Other)9 (3.3)2 (3.8)7 (3.2)Highest completed education (N, %)274No qualification4 (1.5)1 (1.9)3 (1.4)4.79c 0.546 τ = 0.06O Level/GCSE58 (21.2)11 (21.2)47 (21.2)A Level/NVQ27 (9.9)6 (11.5)21 (9.5)Diploma/BTEC42 (15.3)9 (17.3)33 (14.9)University degree84 (30.7)19 (36.5)65 (29.3)Postgraduate degree51 (18.6)6 (11.5)45 (20.3)Other8 (2.9)0 (0.0)8 (3.6)Employment status (N, %)274Paid full-time104 (38.0)29 (55.8)75 (33.8)27.93c < 0.001** V = 0.33Paid part-time82 (29.9)7 (13.5)75 (33.8)Unpaid volunteer work5 (1.8)0 (0.0)5 (2.3)Sick leave2 (0.7)0 (0.0)2 (0.9)Unemployed4 (1.5)0 (0.0)4 (1.8)Student/Pupil0 (0.0)0 (0.0)0 (0.0)Retired36 (13.1)14 (26.9)22 (9.9)Househusband/wife24 (8.8)0 (0.0)24 (10.8)Other17 (6.2)2 (3.8)15 (6.8)Marital status (N, %)274Married199 (72.6)38 (73.1)161 (72.5)2.80c 0.847 V = 0.09In a relationship and cohabiting17 (6.2)5 (9.6)12 (5.4)In a relationship and not cohabiting6 (2.2)1 (1.9)5 (2.3)Single5 (1.8)1 (1.9)4 (1.8)Divorced31 (11.3)4 (7.7)27 (12.2)Separated8 (2.9)1 (1.9)7 (3.2)Widowed8 (2.9)2 (3.8)6 (2.7)IQR: interquartile range; N: number; NA: not applicable; r: Pearson’s correlation coefficient; ϕ: Phi coefficient; τ: Kendall’s tau-c; V: Cramer’s VaMann-Whitney U test b Fisher’s Exact test c Fisher-Freeman-Halton’s Exact test** Statistically significant at the p < 0.01 threshold Table 1Parent baseline demographic characteristicsParent variables N Overall(N = 276)Fathers(N = 52)Mothers(N = 224)Test statisticp-valueEffect size Age 272Median (IQR)54.0 (9)55.0 (14)53.0 (7)4720.5a 0.070 r= − 0.11Ethnicity (N, %)273White264 (96.7)50 (96.2)214 (96.8)NAb 0.682 ϕ= − 0.02Others (Asian, Black, Mixed, Other)9 (3.3)2 (3.8)7 (3.2)Highest completed education (N, %)274No qualification4 (1.5)1 (1.9)3 (1.4)4.79c 0.546 τ = 0.06O Level/GCSE58 (21.2)11 (21.2)47 (21.2)A Level/NVQ27 (9.9)6 (11.5)21 (9.5)Diploma/BTEC42 (15.3)9 (17.3)33 (14.9)University degree84 (30.7)19 (36.5)65 (29.3)Postgraduate degree51 (18.6)6 (11.5)45 (20.3)Other8 (2.9)0 (0.0)8 (3.6)Employment status (N, %)274Paid full-time104 (38.0)29 (55.8)75 (33.8)27.93c < 0.001** V = 0.33Paid part-time82 (29.9)7 (13.5)75 (33.8)Unpaid volunteer work5 (1.8)0 (0.0)5 (2.3)Sick leave2 (0.7)0 (0.0)2 (0.9)Unemployed4 (1.5)0 (0.0)4 (1.8)Student/Pupil0 (0.0)0 (0.0)0 (0.0)Retired36 (13.1)14 (26.9)22 (9.9)Househusband/wife24 (8.8)0 (0.0)24 (10.8)Other17 (6.2)2 (3.8)15 (6.8)Marital status (N, %)274Married199 (72.6)38 (73.1)161 (72.5)2.80c 0.847 V = 0.09In a relationship and cohabiting17 (6.2)5 (9.6)12 (5.4)In a relationship and not cohabiting6 (2.2)1 (1.9)5 (2.3)Single5 (1.8)1 (1.9)4 (1.8)Divorced31 (11.3)4 (7.7)27 (12.2)Separated8 (2.9)1 (1.9)7 (3.2)Widowed8 (2.9)2 (3.8)6 (2.7)IQR: interquartile range; N: number; NA: not applicable; r: Pearson’s correlation coefficient; ϕ: Phi coefficient; τ: Kendall’s tau-c; V: Cramer’s VaMann-Whitney U test b Fisher’s Exact test c Fisher-Freeman-Halton’s Exact test** Statistically significant at the p < 0.01 threshold Parent baseline demographic characteristics IQR: interquartile range; N: number; NA: not applicable; r: Pearson’s correlation coefficient; ϕ: Phi coefficient; τ: Kendall’s tau-c; V: Cramer’s V aMann-Whitney U test b Fisher’s Exact test c Fisher-Freeman-Halton’s Exact test ** Statistically significant at the p < 0.01 threshold Table 2Comparisons of the change patterns in perceived caregiver skills and psychological distress between parents over time (N = 276)Parent variableModel fitConditional R2dfLRT X2p-valueMarginal R2dfLRT X2p-valueF (df, df [res])p-value CASK Scale CASK Total (N = 274)Parent sex3.03 (1, 272) 0.083 Time0.509333.58 < 0.001** 0.02790.73 < 0.001** 0.40 (3, 437) 0.754 Parent sex X Time2.98 (3, 437) 0.031* CASK-Bigger Picture (N = 274)Parent sex0.79 (1, 272) 0.375 Time0.479251.80 < 0.001** 0.02741.55 < 0.001** 2.10 (3, 437) 0.099 Parent sex X Time3.16 (3, 437) 0.024* CASK-Self-care (N = 274)Parent sex4.42 (1, 272) 0.036* Time0.469355.32 < 0.001** 0.05776.87 < 0.001** 2.84 (3, 435) 0.038* Parent sex X Time3.19 (3, 435) 0.023* CASK-Biting-your-tongue (N = 274)Parent sex3.11 (1, 272) 0.079 Time0.529245.88 < 0.001** 0.04772.62 < 0.001** 1.13 (3, 435) 0.336 Parent sex X Time5.15 (3, 435) 0.002** CASK-Insight&Acceptance (N = 274)Parent sex8.79 (1, 272) 0.003** Time0.589342.94 < 0.001** 0.04747.92 < 0.001** 1.16 (3, 435) 0.323 Parent sex X Time1.37 (3, 435) 0.251 CASK-Emotional Intelligence (N = 274)Parent sex0.0006 (1, 272) 0.980 Time0.459221.76 < 0.001** 0.01733.88 < 0.001** 0.86 (3, 435) 0.463 Parent sex X Time1.02 (3, 435) 0.385 CASK-Frustration Tolerance (N = 274)Parent sex0.48 (1, 272) 0.487 Time0.559327.46 < 0.001** 0.01731.73 < 0.001** 1.32 (3, 435) 0.268 Parent sex X Time2.43 (3, 435) 0.065 DASS-21 Scale DASS-21 Total (N = 276)Parent sex0.41 (1, 274) 0.524 Time0.579314.65 < 0.001** 0.01736.48 < 0.001** 0.774(3, 476) 0.509 Parent sex X Time0.58 (3, 476) 0.632 DASS-21-Stress (N = 276)Parent sex0.54 (1, 274) 0.464 Time0.499226.55 < 0.001** 0.01729.91 < 0.001** 1.19 (3, 475) 0.313 Parent sex X Time0.72 (3, 475) 0.537 DASS-21-Anxiety (N = 276) Parent sex0.32 (1, 274) 0.569 Time0.549261.20 < 0.001** 0.01712.410.0880.80 (3, 475) 0.494 Parent sex X Time0.12 (3, 475) 0.951 DASS-21-Depression (N = 276) Parent sex0.17 (1, 274) 0.681 Time0.599301.26 < 0.001** 0.01720.860.004**0.29 (3, 476) 0.836 Parent sex X Time0.92 (3, 476) 0.428 Model equations: Each CASK or DASS-21 score ~ 1 + Parent sex + Time + Parent sex X Time + (1|Participant). Parent sex: Fathers versus Mothers; Time: Baseline, Month 6, Month 12, Month 18ASK: Caregiver Skills Scale; DASS-21: Depression, Anxiety, and Stress Scale-Short form; df: degree of freedom; LRT X2: Likelihood Ratio Test-Chi-square; N: number; res: residual**Significant at the p < 0.01 threshold; *Significant at the p < 0.05 threshold Table 2Comparisons of the change patterns in perceived caregiver skills and psychological distress between parents over time (N = 276)Parent variableModel fitConditional R2dfLRT X2p-valueMarginal R2dfLRT X2p-valueF (df, df [res])p-value CASK Scale CASK Total (N = 274)Parent sex3.03 (1, 272) 0.083 Time0.509333.58 < 0.001** 0.02790.73 < 0.001** 0.40 (3, 437) 0.754 Parent sex X Time2.98 (3, 437) 0.031* CASK-Bigger Picture (N = 274)Parent sex0.79 (1, 272) 0.375 Time0.479251.80 < 0.001** 0.02741.55 < 0.001** 2.10 (3, 437) 0.099 Parent sex X Time3.16 (3, 437) 0.024* CASK-Self-care (N = 274)Parent sex4.42 (1, 272) 0.036* Time0.469355.32 < 0.001** 0.05776.87 < 0.001** 2.84 (3, 435) 0.038* Parent sex X Time3.19 (3, 435) 0.023* CASK-Biting-your-tongue (N = 274)Parent sex3.11 (1, 272) 0.079 Time0.529245.88 < 0.001** 0.04772.62 < 0.001** 1.13 (3, 435) 0.336 Parent sex X Time5.15 (3, 435) 0.002** CASK-Insight&Acceptance (N = 274)Parent sex8.79 (1, 272) 0.003** Time0.589342.94 < 0.001** 0.04747.92 < 0.001** 1.16 (3, 435) 0.323 Parent sex X Time1.37 (3, 435) 0.251 CASK-Emotional Intelligence (N = 274)Parent sex0.0006 (1, 272) 0.980 Time0.459221.76 < 0.001** 0.01733.88 < 0.001** 0.86 (3, 435) 0.463 Parent sex X Time1.02 (3, 435) 0.385 CASK-Frustration Tolerance (N = 274)Parent sex0.48 (1, 272) 0.487 Time0.559327.46 < 0.001** 0.01731.73 < 0.001** 1.32 (3, 435) 0.268 Parent sex X Time2.43 (3, 435) 0.065 DASS-21 Scale DASS-21 Total (N = 276)Parent sex0.41 (1, 274) 0.524 Time0.579314.65 < 0.001** 0.01736.48 < 0.001** 0.774(3, 476) 0.509 Parent sex X Time0.58 (3, 476) 0.632 DASS-21-Stress (N = 276)Parent sex0.54 (1, 274) 0.464 Time0.499226.55 < 0.001** 0.01729.91 < 0.001** 1.19 (3, 475) 0.313 Parent sex X Time0.72 (3, 475) 0.537 DASS-21-Anxiety (N = 276) Parent sex0.32 (1, 274) 0.569 Time0.549261.20 < 0.001** 0.01712.410.0880.80 (3, 475) 0.494 Parent sex X Time0.12 (3, 475) 0.951 DASS-21-Depression (N = 276) Parent sex0.17 (1, 274) 0.681 Time0.599301.26 < 0.001** 0.01720.860.004**0.29 (3, 476) 0.836 Parent sex X Time0.92 (3, 476) 0.428 Model equations: Each CASK or DASS-21 score ~ 1 + Parent sex + Time + Parent sex X Time + (1|Participant). Parent sex: Fathers versus Mothers; Time: Baseline, Month 6, Month 12, Month 18ASK: Caregiver Skills Scale; DASS-21: Depression, Anxiety, and Stress Scale-Short form; df: degree of freedom; LRT X2: Likelihood Ratio Test-Chi-square; N: number; res: residual**Significant at the p < 0.01 threshold; *Significant at the p < 0.05 threshold Comparisons of the change patterns in perceived caregiver skills and psychological distress between parents over time (N = 276) Model equations: Each CASK or DASS-21 score ~ 1 + Parent sex + Time + Parent sex X Time + (1|Participant). Parent sex: Fathers versus Mothers; Time: Baseline, Month 6, Month 12, Month 18 ASK: Caregiver Skills Scale; DASS-21: Depression, Anxiety, and Stress Scale-Short form; df: degree of freedom; LRT X2: Likelihood Ratio Test-Chi-square; N: number; res: residual **Significant at the p < 0.01 threshold; *Significant at the p < 0.05 threshold Pearson’s correlation coefficient analyses were conducted to identify the correlation matrix of repeated measures in each model. Depending on the consistency of the correlation between the timepoints, “compound symmetry” or “autoregressive of order 1 (i.e., AR [1])” were used as the residual structures to model the correlation between repeated measurements within a subject. Compound symmetry was employed if the correlation was similar across pairs of the measurements. In contrast, if the correlation generally decreased as the time lags increased between the measurements, AR(1) was used to account for the dependency between the residuals of the model. Improved model fit indices further justified the residuals structure preferences, with lower Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values. The significance level of p < 0.05 was used across the analyses. A multiple testing correction was not applied to the results of the LMM analyses because of the exploratory nature of this study [41, 42]. However, the pairwise comparisons were adjusted using the Bonferroni correction [43]. Standardised regression coefficients (β) and standardised effect sizes were calculated by dividing regression coefficients or the mean differences between the comparison groups by the standard deviation of the outcome measured at baseline [44, 45]. Following Cohen’s (1969) [46] convention, effect sizes were interpreted as small (≥ |0.20|), medium (≥ |0.50|), and large (≥ |0.80|). Results Preliminary analyses Baseline characteristics Parent baseline characteristics Baseline demographic characteristics were compared between the trial arms per parent group (i.e., fathers in ECHOMANTRA vs. TAU, mothers in ECHOMANTRA vs. TAU). No significant differences were found in age, ethnicity, highest level of education completed, marital and employment status between TAU and ECHOMANTRA for the father or mother groups (all p > 0.05 and small-to-medium effect sizes; see Additional file 2). Similarly, LMM analyses were conducted in each parent sample to assess whether the CASK and DASS-21 scores significantly differed between the trial arms over time. The ECHOMANTRA and TAU groups exhibited similar trajectories of change in perceived carer skills and psychological distress for both fathers or mothers (all p > 0.05 for the trial arm allocation x time interaction term; see Additional files 6 and 7). Therefore, the further analyses were performed using both trial arms combined (i.e., did not differentiate between trial arms). Patient baseline characteristics There were no significant differences in the patients’ demographic (i.e., age, sex) or clinical (i.e., AN treatment type, years with eating disorder, history of treatment under the MHA, number of treatments under the MHA, BMI, the DASS-21 and EDE-Q total scores) characteristics at baseline between those who nominated their father versus mother as their primary caregiver in the TRIANGLE trial (all p > 0.05 and small-to-medium effect sizes; see Additional file 3). Primary analyses Baseline characteristics Parent baseline characteristics More mothers than fathers (81.2% vs. 18.8%) were nominated by the patients as the primary carers. No significant differences were found in the sociodemographic characteristics between fathers and mothers (all p > 0.05; Table 1), except for the employment status. The median age of fathers was 55.0 (IQR = 14), whereas the median age of mothers was 53.0 (IQR = 7) (U = 4720.5, p = 0.070, r= −0.11). The majority of both parent groups self-identified as White (96.2% of the fathers; 96.8% of the mothers; p = 0.682, ϕ= −0.02), married (73.1% of the fathers; 72.5% of the mothers; p = 0.847, Cramer’s V = 0.09), and had a diploma or above (65.3% of the fathers; 64.5% of the mothers; p = 0.546, τ = 0.06). There was a significant association between parent sex and employment status (p < 0.001, Cramer’s V = 0.33). While fathers were most commonly in paid full- or part-time (55.8% and 13.5%, respectively) employment or retired (26.9%), mothers showed greater diversity in their employment status, with paid full- or part-time (both 33.8%), unpaid domestic roles (10.8%), retired (9.9%), and other employments (6.8%). As employment status may conceivably have an effect on the dependent variables of perceived caregiver skills and psychological distress, it was henceforth included in sensitivity analyses for the main LMM analyses as a two-level categorical variable (i.e., ‘paid employment’ and ‘others’) in order to minimise model overfitting. Patient baseline characteristics In the patient-parent dyads group, most patients were females (92.8%) and in early adulthood (median age = 20.84, IQR = 6). The median AN duration was 5 years (IQR = 7), and the average BMI at admission was 14.5 kg/m2 (SD = 1.95). About 60% of patients was not treated under the MHA. The majority of patients were recruited while being admitted for inpatient care (79.7%). At baseline, there were no significant differences between the inpatient and day patient groups in age, self-reported sex, history of treatment under the MHA, number of treatments under the MHA, EDE-Q or DASS-21 total scores (all p > 0.05), except for illness duration and BMI. The median AN duration in the inpatient group (median = 5, IQR = 7) was significantly longer than in the day patient group (median = 3, IQR = 5) (U = 4784.0, p = 0.016, r= −0.15). Conversely, the average BMI was significantly lower in the inpatient group (M = 14.25 kg/m2, SD = 1.94) than in the day patient group (M = 15.46 kg/m2, SD = 1.65) (t= −4.29, p < 0.001, d= −0.64, 95% CIs [−0.940, −0.343]). For more detailed information about the patient baseline characteristics, see Additional files 3 and 4. Comparisons of longitudinal changes in perceived caregiver skills between fathers and mothers Model convergence was achieved in all LMM analyses, with no estimation problems observed. The interaction effect between parental sex and time was significant in the models for the CASK total score and three subscales “Bigger Picture”, “Self-care”, and “Biting-your-tongue” (total score, F (3, 437) = 2.98, p = 0.031; “Bigger Picture”, F (3, 437) = 3.16, p = 0.024; “Self-Care”, F (3, 435) = 3.19, p = 0.023; “Biting-your-tongue”, F (3, 435) = 5.15, p = 0.002; Table 2), indicating that the mean changes in these carer skills significantly differentiated between mothers and fathers over time. The divergences in the CASK change scores between fathers and mothers were apparent only from baseline to month 18, but not from baseline to month 6 or 12 (Additional file 8 and Fig. 2a–d). The main effect of parent sex was significant only for the subscales of “Self-Care” (F (1, 272) = 4.42, p = 0.036) and “Insight and Acceptance” (F (1, 272) = 8.79, p = 0.003; Table 2; Fig. 2c and e). Time was significant only for the “Self-Care” subscale (F (3, 435) = 2.84, p = 0.038; Table 2). There were no interaction or main effects for the subscales of “Emotional Intelligence” and “Frustration Tolerance” (Table 2; Fig. 2f and g). The analysis results did not change in any models when including parent employment status. Overall, fathers and mothers exhibited diverging trajectories over time on the CASK total and the “Bigger Picture”, “Self-Care”, and “Biting-your-tongue” subscales, but not on the “Insight and Acceptance”, “Emotional Intelligence”, and “Frustration Tolerance” subscales. These divergences between the groups became apparent at month 18, but not at month 6 or 12. The main effects of parent sex and time were evident only for the “Self-Care” and “Insight and Acceptance” subscales. The raw means and the estimated marginal means were presented in Additional files 9 and 10. Overall CASK scores Post-hoc tests revealed that there were no differences between fathers and mothers in overall CASK scores at baseline, 6-, 12-, and 18-months (all p > 0.05 after Bonferroni correction; Additional file 11). However, the average change in the CASK total scores from baseline to month 18 was significantly greater for mothers (B = 260.08, β = 0.65, SE = 90.5, t = 2.87, p = 0.004, 95% CIs = [82.2, 438.0]; Additional file 8). Fathers did not show a significant change in the CASK total scores from baseline to 6-, 12-, and 18-months (all pBonferroni > 0.05; Additional file 11). In contrast, mothers demonstrated a significant improvement in their skills from baseline to 18-month (mean difference = −142.27, SE = 39.0, t = −3.65, pBonferroni = 0.008; Additional file 11 and Fig. 2a), although this was not significant from baseline to 6- or 12-months. Bigger picture subscale There were similarly no differences between fathers and mothers at baseline, 6-, 12-, and 18-months for the “Bigger Picture” subscale (all p > 0.05 after Bonferroni correction; Additional file 12). From baseline to 18-month (but not 6- or 12-months), the average change in mothers’ ability to adopt a positive long-term perspective regarding their child’s illness journey was significantly higher compared to fathers (B = 11.67, β = 0.65, SE = 3.85, t = 3.03, p = 0.003, 95% CIs = [4.10, 19.24]; Additional file 8). After Bonferroni correction, there were no differences between baseline and 6-, 12-, and 18-months in this subscale for either fathers or mothers (all pBonferroni > 0.05; Additional file 12); before correction, a similar trajectory was found to the Bigger Picture score whereby only mothers showed an improvement from baseline to month 18 (mean difference = −3.37, SE = 1.66, t = −2.03, p = 0.043; Additional file 12 and Fig. 2b). Self-care subscale Fathers had a significantly higher self-care score than mothers at baseline (mean difference = 11.81, SE = 3.51, t = 3.36, pBonferroni = 0.025), but not at the other time-points (all pBonferroni > 0.05; Additional file 13). Additionally, the average change in this subscale from baseline to month 18 was significantly greater for mothers (B = 12.19, β = 0.56, SE = 4.54, t = 2.69, p = 0.008, 95% CIs = [3.27, 21.11]; Additional file 8) whereby mothers (but not fathers) showed an improvement in self-care skill at month 18 (mean difference = −12.01, SE = 1.94, t = −6.17, pBonferroni < 0.001; Additional file 13 and Fig. 2c). Concerning the main effect of parental sex, on average, mothers reported lower self-care than fathers across time (B = −6.91, β = −0.31, SE = 3.28 t = −2.10, p = 0.04, 95% CIs = [−13.37, −0.44]; Additional file 8). Regarding the main effect of time, on average, parents showed more improvement in self-care skill at months 12 (B = 5.19, β = 0.24, SE = 2.10, t = 2.48, p = 0.014, 95% CIs = [1.07, 9.32]) and 18 (B = 5.91, β = 0.27, SE = 2.27, t = 2.60, p = 0.010, 95% CIs = [1.45, 10.37]; Additional file 8), not month 6. However, post-hoc tests did not survive Bonferroni correction (all pBonferroni > 0.05; Additional file 14). Biting-your-tongue subscale For this subscale, fathers had significantly higher average score than mothers at baseline (mean difference = 11.51, SE = 3.53, t = 3.26, pBonferroni= 0.035), but not at the other time-points (all pBonferroni > 0.05; Additional file 15). Again, the average change in this caregiver skill from baseline to month 18 was significantly greater for mothers (B = 17.60, β = 0.83, SE = 4.65, t = 3.79, p < 0.001, 95% CIs= [8.46, 26.74]; Additional file 8) whereby mothers, but not fathers, showed an improvement in controlling their urge to enquire repetitive conversations with their child with AN at month 18 (mean difference = −12.21, SE = 2.01, t = −6.08, pBonferroni < 0.001; Additional file 15 and Fig. 2d). Figs. 2a–d Profile plot displaying the changes in estimated marginal means (errors bars indicate 95% CIs) over time by parent sex for the CASK total, Bigger Picture, Self-Care, and Biting-your-tongue scores. e–g Profile plot displaying the changes in estimated marginal means (errors bars indicate 95% CIs) over time by parent sex for the Insight and Acceptance, Emotional Intelligence, and Frustration Tolerance scores a–d Profile plot displaying the changes in estimated marginal means (errors bars indicate 95% CIs) over time by parent sex for the CASK total, Bigger Picture, Self-Care, and Biting-your-tongue scores. e–g Profile plot displaying the changes in estimated marginal means (errors bars indicate 95% CIs) over time by parent sex for the Insight and Acceptance, Emotional Intelligence, and Frustration Tolerance scores Insight and acceptance subscale The analysis examining the main effect of parental sex on Insight and Acceptance subscale scores found that, on average, mothers had lower scores in the tolerance and management of unpleasant emotions than fathers across time (B = −9.55, β = −0.43, SE = 3.22, t = −2.97, p = 0.003, 95% CIs= [−15.89, −3.21]; Additional file 8 and Fig. 2e). Comparisons of longitudinal changes in psychological distress between fathers and mothers No convergence errors were observed in any LMM analyses. The analysis results did not change in any models when including parent employment status. The raw means and the estimated marginal means were given in Additional files 9 and 10. Although visual inspection suggested slight downward trends in mothers’ distress scores compared to fathers (Additional file 10 and Fig. 3a–d), there were no significant parent sex x time interaction effects or main effects of parental sex or time found for the overall DASS-21 or any of its subscales (all p > 0.05; Table 2). These findings indicate that the trajectories of change in depression, anxiety, and stress levels were similar for fathers and mothers over time, meaning that the average changes in the DASS-21 total and subscale scores from baseline to months 6, 12, or 18 did not significantly differ between fathers and mothers. Fig. 3a–d Profile plot displaying changes in estimated marginal means (errors bars indicate 95% CIs) over time by parent sex for the DASS-21 total, Depression, Anxiety, and Stress scores a–d Profile plot displaying changes in estimated marginal means (errors bars indicate 95% CIs) over time by parent sex for the DASS-21 total, Depression, Anxiety, and Stress scores Discussion This study aimed to examine the 18-month trajectories of change and differences at baseline, month 6, 12, and 18 in perceived caregiver skills (CASK) and psychological distress (DASS-21) in patient-selected father and mother carers who took part in the TRIANGLE trial. Fathers and mothers did not show significant differences in the CASK scores at any time points, except for higher paternal “Self-Care” and “Biting-your-tongue” skills at baseline. The trajectories of change differed between the groups over the 18-month period for the overall CASK and for the “Bigger Picture”, “Self-Care”, and “Biting-your-tongue” subscales. After Bonferroni correction, mothers showed significant improvements in the overall CASK and the “Self-Care” and “Biting-your-tongue” skills from baseline to month 18, whereas fathers did not show significant changes over 18 months. Nevertheless, fathers had significantly higher scores in “Self-Care” and “Insight and Acceptance” across time. No significant results were found for the “Emotional Intelligence” and “Frustration Tolerance” skills, or for any of the DASS-21 scores (i.e., the total, depression, anxiety, and stress scores). These findings indicate that mothers’ caring skills improved over the study period. A possible explanation is mothers’ engagement with informal (e.g., skills trainings, carer manuals, workshops) and peer support resources during their child’s transition. For example, a recent qualitative study involving mainly mothers of adults with AN examined carers’ needs and experiences during the transition process [13]. The findings revealed that many mothers sought various forms of support (e.g., reading psychoeducational books, individual therapy, participation in hobbies, attending peer support groups) to cope with the caregiving role and to help sustain their child’s treatment gains post-discharge, and it is possible that these contributed to the improvement in their skills. In the current study, mothers may likewise have used informal support resources more than fathers during the study period, which could have contributed to improvements their caregiving skills, such as “Self-Care” or “Biting-your-tongue”. However, no data were collected on carers’ informal support resources in the primary trial [19]. Additionally, because of the limited number of fathers in qualitative studies, no detailed information was provided on whether or how fathers benefited from available support options during their adult child’s transition [14, 15]. These limitations make it difficult to interpret the present findings in light of potential differences between fathers and mothers, highlighting the need for further research to examine how similarities and differences in the use of support resources may influence changes in parents’ caregiving abilities during the transition process. Furthermore, the CASK is a subjective measure; therefore, improvements may reflect changes only in mothers’ perceptions of their caring skills. They, for instance, may have over time developed self-compassion towards themselves as parents and/or adopted more realistic expectations about their caregiving role. Semi-structured interviews alongside subjective assessments of caring abilities could provide a more comprehensive understanding of how perceived competence in caregiving is experienced and expressed among fathers and mothers. No significant changes were observed in fathers’ caring skills over the 18-month period. This may reflect their higher baseline scores in certain aspects of caregiving (i.e., “Self-Care” and “Biting-your-tongue”). In addition, previous studies have highlighted that male carers often require better guidance and signposting to available support when caring for their loved ones during and after intensive treatment [14, 15, 47]. Fathers who served as primary caregivers in the present study may have had limited opportunities to access such support and to further develop their caregiving competence over the course of the study. These factors might explain the relative stability or the non-significant decreases observed in their self-ratings of caring abilities. Alternatively, the findings may suggest that fathers were able to preserve their perceived confidence in skilful caregiving to some extent during their child’s transition. Nevertheless, given the small sample size of fathers and the resulting limited power to detect within-group effects reliably [48], the findings should be interpreted with caution. Further quantitative and qualitative research is warranted to confirm these results and to better understand fathers’ caregiving experiences and associated factors during the transition process, ultimately informing carer interventions that are more attuned to fathers’ unique needs. The trajectories of change in the “Insight and Acceptance”, “Emotional Intelligence”, and “Frustration Tolerance” caregiving abilities did not differ between fathers and mothers over 18 months. Furthermore, no significant main effect of time was observed for these subscales, suggesting relative stability in parents’ scores over time. These similarities may reflect shared parental experiences in these aspects of caregiving during their child’s transition. These skills represent more adaptive ways of unhelpful emotional and behavioural reactions (e.g., high expressed emotions, accommodating behaviours) to the illness commonly exhibited by carers in the management of EDs [31]—such as accepting the patient’s illness, discussing and managing carer’s own feelings, upholding boundaries and rules consistently, and remaining calm, compassionate, and understanding when responding to difficult ED behaviours. Previous research has highlighted that the complex and chronic nature of AN and the associated caregiving strain often make it difficult for carers to sustain these more supportive approaches towards their loved ones [8, 23, 49–51]. In a qualitative study [14], for example, most mothers noticed signs of eating disorder thoughts and behaviours in their adult children during the transition and felt confused about the nature of AN and how to respond. Since both fathers and mothers were the primary carers for their adult child in the present study, they were likely to have encountered similar caregiving experiences and might have found it difficult to enhance their reflective, emotionally attuned, and balanced caregiving capacities over the study period, given the severity of their child’s illness (i.e., requiring intensive treatment) and the potential transition challenges (e.g., insufficient carer support, changes in caregiving role). This may explain why their trajectories of change on the CASK subscales of interest were comparable, with scores remaining broadly stable over time. Nonetheless, it should be noted that there was an imbalance between the sample sizes of fathers and mothers, meaning that the non-significant time effect likely reflects stability primarily among mothers rather than equally across both parents. Further longitudinal research with more balanced samples is needed to corroborate these findings and to examine how changes in the child’s AN symptoms and transition-related difficulties may influence both parents’ perceptions of caregiving strain and competence. Interestingly, distress scores for both fathers and mothers were within the normal-to-mild clinical range across all subscales at baseline, suggesting broadly similar initial levels of distress. This finding differs from previous quantitative and qualitative research reporting elevated distress for many parents of individuals with AN [4, 22, 24, 27, 51]. However, variations in patient characteristics (e.g., age, treatment status, illness duration) and how ‘distress’ is defined and/or measured across studies may explain this discrepancy. Given the relatively longer illness duration in the present sample, it is plausible that several patients had already undergone inpatient treatment prior to their inclusion in the larger trial [19]. Such experience may have contributed to parents’ relatively lower distress at baseline, possibly through familiarity with the treatment process. For instance, Raenker et al. [3] reported distress levels within a similar clinical severity range among parents of patients with comparable characteristics (e.g., age, need for intensive care, years with AN). Nevertheless, our interpretation remains tentative due to the lack of data on prior admissions in the primary research [19]. Moreover, the trajectories of depression, anxiety, and stress scores did not significantly change over the 18-month period between fathers and mothers. Our findings contrast with cross-sectional studies involving parents of adolescents with AN, in which mothers reported higher distress than fathers [5, 23, 25]. This discrepancy could be attributed to the fact that mothers, who are often the primary carers of their adolescent child, tend to exhibit higher levels of factors (e.g., emotional overinvolvement, accommodating behaviours) closely associated with psychological distress compared to fathers, who are mostly secondary carers [23, 24, 28]. However, these aspects of parental caregiving were not explored in the larger trial [19], making it difficult to interpret our results within this framework. Given that both parent groups were the primary caregivers of adults with AN in the present study, and that fathers and mothers reported relatively low distress scores at baseline, it seems reasonable that they showed comparable change trajectories of distress over time. A further explanation may be related to carers’ adaptive responses to the prolonged course of AN. In a qualitative study exploring the experiences of carers of adults with EDs [4], some fathers and mothers reported having come to accept recovery as a gradual process and having learned to live with the situation over time. In the current study, parents might have developed greater adaptability in managing their caregiving role throughout their child’s relatively long illness course (~ 5 years), allowing them to adjust more effectively to caregiving strain during the transition. This may, in turn, have contributed to distress levels within the normal to mild clinical ranges in both parent groups across the study period. Nevertheless, future research is needed to examine the long-term course of psychological distress among fathers and mothers of adult patients, and how changes over time relate to parents’ perceptions of their caregiving skills. Strengths and limitations To our knowledge, this is the first longitudinal study exploring the changes in perceived caregiver skills and psychological distress in fathers and mothers of adult patients with AN requiring intensive care. This is important for several reasons. Firstly, previous research has largely focused on parents of adolescents, with an emphasis on carer strain and associated factors, mainly employing a cross-sectional design [5, 23]. Secondly, although a limited number of longitudinal studies have examined parental caregiving and/or depressive symptoms [24, 52], these studies did not investigate the comparison of the patterns of changes in skilful caregiving and/or psychological distress between parents of adults with AN. Another strength of this study is the inclusion of fathers. Irrespective of the ED diagnosis and/or patient age, fathers are often perceived as secondary carers compared to mothers, even if they are significant supporters of their children with an ED [26, 28]. In the present study, fathers represented approximately 20% of the parent sample. This proportion is comparable to European statistics on the gender distribution of family carers (15% male versus 85% female) [53]. However, we could not find any study reporting the proportions of fathers and mothers caring for their children with a physical or mental health condition as their primary caregivers. This makes it difficult to directly compare the percentage of fathers in our study with global statistics. Additionally, the reasons why fewer patients nominated fathers as their primary carer in the present study remain unclear. Further research is needed to investigate parental primary caregiving proportions in EDs and to identify the factors that may be associated with these proportions. This study has also some limitations. Notably, due to the research design of the larger trial [19], only one primary caregiver per patient was included in the study. As fathers and mothers were not the parents of the same patients, it was not possible to make direct comparisons between parent groups or to explore potential congruences or divergences in the parental outcomes [52, 54]. Moreover, we examined only group differences and changes in carer skills in relation to our research aims. However, the study by Salerno et al. [52] found that the skilful caring behaviours of parents at baseline were related to the rate of change of the patient’s AN symptoms in one year. Therefore, it would be interesting to explore how distinct change patterns in the caregiving skills of fathers and mothers relate to patient outcomes over time. It is also noteworthy that there was an imbalance between the sample sizes of parents, accompanied by increasing attrition rates over the follow-up period. Although such problems are common in longitudinal research, and robust statistical methods (i.e., linear mixed-effects models) were used to address missing data [38, 55], the results may have been biased or meaningful differences between fathers and mothers might not have been detected. This indicates the need for future studies with more balanced group sizes to validate the findings of this study. Finally, most parents self-identified as White and the majority of patients were females with AN necessitating intensive hospital care. This limits the generalisability of the study findings to more diverse parental populations, including those from different ethnic backgrounds and those of adolescents or males with other ED subtypes at varying stages of the illness. Clinical implications The findings of the present study have various clinical implications. The existing interventions (e.g., Family-based Therapy, skills-based trainings) aim to improve parents’ caring abilities, but they generally do not consider the potential differences between fathers and mothers. However, our findings revealed that changes in several parental caregiving skills followed divergent trajectories over time. Additionally, the changes were more apparent at 18 months, meaning that they took a significant amount of time to benefit their child. These may indicate the importance of developing long-term carer interventions that address paternal and maternal needs across different stages of the child’s transition. Fathers and mothers also showed comparable trajectories of change in several caring abilities (i.e., “Insight and Acceptance”, “Emotional Intelligence”, and Frustration Tolerance”) over the 18-month period. This may suggest that gender-specific tailoring of carer interventions is not necessary when supporting parents in developing reflective and emotionally attuned caregiving skills during the transition process. In addition, parents’ scores on these subscales remained broadly stable over time. While this finding should be interpreted cautiously due to the aforementioned limitation (i.e., imbalance in group sizes), it may nonetheless indicate the need for support programmes that focus on strengthening parents’ emotion regulation abilities. Such support could help parents sustain emotionally grounded and balanced caregiving and reduce less adaptive behavioural responses when faced with emotionally challenging situations during their child’s transition. Moreover, the results of this study can be utilised to support parents in a more empowering manner. Although carer interventions predominantly target unhelpful parental behaviours (e.g., accommodating and enabling behaviours, high levels of expressed emotions), the skilful dimensions of caregiving are often disregarded. In our study, for example, fathers consistently scored higher than mothers over time on their ability to tolerate unpleasant emotions (“Insight and Acceptance”) and to take time for themselves and other family members (“Self-Care”). Highlighting these strengths may help alleviate feelings of blame and isolation that fathers usually experience in relation to their engagement with their child’s treatment. Thus, focusing on the skilful aspects of caregiving as well as difficulties might enhance parents’ self-efficacy and reduce distress, which in turn could positively influence patient outcomes [26, 56]. The final implication concerns the need for aftercare support for parents of individuals with AN. Recent systematic reviews have demonstrated that post-intensive care interventions mostly targeted patients to support their transitions from the hospital, rather than interventions for their carers (e.g., parents) [16, 17]. While parental involvement in the treatment of adults with EDs is generally more limited compared to that of adolescents because of confidentiality policies or an exclusion from service provision [4, 15], 80% of the total carer sample in the primary research consisted of parents (vs. 20% non-parental relationship; [19]). This highlights that parents continue to take on the primary caregiving role for adult patients, underlining the need for more improved aftercare interventions to facilitate the transition process for parents. Conclusions This exploratory mixed-design (i.e., within-subjects and between-groups) secondary analysis of data from the TRIANGLE trial is the first longitudinal study comparing the changes in perceived caregiver skills and psychological distress between chosen fathers and mothers of adult AN patients over 18 months during the transition from intensive hospital care. Although parents did not show significant differences in any CASK outcome at any time point (except for higher paternal scores in “Self-Care” and “Biting-your-tongue” [control the urge for repeated conversations with the patient] at baseline), the trajectories of change differed in the overall CASK, the “Bigger Picture (adopting a long-term perspective to their child’s recovery)”, “Self-Care”, and “Biting-your-tongue” subscales between the groups over 18 months. After Bonferroni correction, mothers showed significant improvements in the overall CASK and the “Self-Care” and “Biting-your-tongue” skills from baseline to month 18, while no significant changes were observed for fathers in this period. Nonetheless, fathers had significantly higher average scores in “Self-care” and “Insight and Acceptance” (the tolerance and management of unpleasant emotions) across time. The changes in the “Insight and Acceptance”, “Emotional Intelligence (discussing and managing feelings)”, and “Frustration Tolerance (resolving conflicts with a firm and understanding manner)” skills and psychological distress levels (i.e., the total, depression, anxiety, and stress scores) followed similar trajectories in fathers and mothers over the study period. These findings suggest that fathers and mothers might have distinct needs to enhance their caregiving skills during their adult child’s transition from an intensive treatment to the community. Therefore, further quantitative and qualitative research is warranted to explore parents’ needs, and long-term carer support appears to be essential during this process. Supplementary Information Additional file 1: Box 1 Information about the trial arms in the TRIANGLE study. Additional file 1: Box 1 Information about the trial arms in the TRIANGLE study. Additional file 2: Table S1 Parent baseline demographics split by the trial arms. Additional file 2: Table S1 Parent baseline demographics split by the trial arms. Additional file 3: Table S2 Comparisons of patient demographic and clinical characteristics at baseline between those who had a father or mother as the primary carer. Additional file 3: Table S2 Comparisons of patient demographic and clinical characteristics at baseline between those who had a father or mother as the primary carer. Additional file 4: Table S3 Comparisons of the baseline demographic and clinical characteristics between the inpatient and day patient groups. Additional file 4: Table S3 Comparisons of the baseline demographic and clinical characteristics between the inpatient and day patient groups. Additional file 5: Table S4 Missing data. Additional file 5: Table S4 Missing data. Additional file 6: Table S5a Comparisons of the change patterns in perceived caregiver skills and psychological distress of fathers between trial arms over time. Additional file 6: Table S5a Comparisons of the change patterns in perceived caregiver skills and psychological distress of fathers between trial arms over time. Additional file 7: Table S5b Comparisons of the change patterns in perceived caregiver skills and psychological distress of mothers between trial arms over time. Additional file 7: Table S5b Comparisons of the change patterns in perceived caregiver skills and psychological distress of mothers between trial arms over time. Additional file 8: Table S6 Coefficients for the main and interaction effects on perceived caregiver skills and psychological distress. Additional file 8: Table S6 Coefficients for the main and interaction effects on perceived caregiver skills and psychological distress. Additional file 9: Table S7a Raw means for the CASK and DASS-21 total and subscale scores. Additional file 9: Table S7a Raw means for the CASK and DASS-21 total and subscale scores. Additional file 10: Table S7b Estimated marginal means for the interaction effect between parent sex and time on the CASK and DASS-21 total and subscale scores. Additional file 10: Table S7b Estimated marginal means for the interaction effect between parent sex and time on the CASK and DASS-21 total and subscale scores. Additional file 11: Table S8 Pairwise comparisons for the interaction effect between parent sex and time on the CASK total scores. Additional file 11: Table S8 Pairwise comparisons for the interaction effect between parent sex and time on the CASK total scores. Additional file 12: Table S9 Pairwise comparisons for the interaction effect between parent sex and time on the CASK Bigger Picture subscale scores. Additional file 12: Table S9 Pairwise comparisons for the interaction effect between parent sex and time on the CASK Bigger Picture subscale scores. Additional file 13: Table 10a Pairwise comparisons for the interaction effect between parent sex and time on the CASK Self-Care subscale scores. Additional file 13: Table 10a Pairwise comparisons for the interaction effect between parent sex and time on the CASK Self-Care subscale scores. Additional file 14: Table 10b Pairwise comparisons for the main effect of time on the CASK Self-Care subscale scores. Additional file 14: Table 10b Pairwise comparisons for the main effect of time on the CASK Self-Care subscale scores. Additional file 15: Table S11 Pairwise comparisons for the interaction effect between parent sex and time on the CASK Biting-your-tongue subscale scores. Additional file 15: Table S11 Pairwise comparisons for the interaction effect between parent sex and time on the CASK Biting-your-tongue subscale scores.