Psychometric properties of the German version of the fears of compassion scales

Institute of Psychiatric and Psychosomatic Psychotherapy, Central Institute of Mental Health Mannheim, Mannheim, Germany McLean Hospital Harvard Medical School, Boston, Massachusetts, USA School of Sciences, University of Derby, Derby, UK Clinical for Psychosomatic & Psychotherapy, Schön Clinic Bad Bramstedt, Bad Bramstedt, Germany Faculty of Health, Department of Psychology and Psychotherapy, University Witten/Herdecke, Witten, Germany Clinic for Psychiatry, Psychotherapy and Preventive Medecine, LWL University Hospital, Ruhr University Bochum, Bochum, Germany Department of Public Health and Health Education, University of Education, Freiburg, Freiburg im Breisgau, Germany


| INTRODUCTION
Compassion is a psychological concept that has received increasing scientific interest during the last 20 years. Some define compassion as an emotion (Goetz, Keltner, & Simon-Thomas, 2010), and others define it as a multidimensional construct (Jazaieri et al., ;Strauss et al., 2016). One of the most influential and frequently used definitions is that of Gilbert (2014), who defines compassion as a motif, involving the 'sensitivity to suffering in self and others with a commitment to try alleviate or prevent it' . An increasing number of studies have demonstrated that compassion influences emotional processing, that is, attending to, processing, remembering and reacting to emotional stimuli (Kirby, Doty, Petrocchi, & Gilbert, 2017;Seppälä et al., 2017).
These key processes are directly linked to the activity of the autonomic sympathetic nervous system, which enables emotion-related action tendencies such as the approach to relevant others and caregiving. The activity of the parasympathetic nervous system enables the corresponding calming and soothing tendencies. Previous studies have shown that giving and receiving compassion is physiologically linked to adaptive heart rate variability (e.g., Cosley, McCoy, Saslow, & Epel, 2010;Kim et al., 2020;Kirby et al., 2017;Matos et al., 2017;Petrocchi, Ottaviani, & Couyoumdjian, 2017;Rockliff, Gilbert, McEwan, Lightman, & Glover, 2008), blood pressure and cortisol reactivity (Cosley et al., 2010). Additionally, previous research has shown that compassion training affects the activation of the amygdala and of other brain areas involved in emotional processing and empathy (Derntl et al., 2010;Desbordes et al., 2012;Klimecki, Leiberg, Lamm, & Singer, 2013). Investigations on functional brain plasticity after compassion and empathy training suggest compassion may reflect a new coping strategy to reverse empathic distress and to strengthen resilience (Klimecki et al., 2013;Klimecki, Leiberg, Ricard, & Singer, 2014).
Thus, compassion has recently become the focus of interventions for a range of mental health problems. To date, six empirically based interventions that aim to cultivate compassion have been developed (Kirby, 2017): compassion focused therapy (CFT; Gilbert, 2014), mindful self-compassion (MSC; K. D. Neff & Germer, 2013), compassion cultivation training (CCT; Jinpa, 2010); cognitively based compassion training (CBCT; Pace et al., 2009), cultivating emotional balance (CEB; Kemeny et al., 2012) and loving-kindness (LKM) or compassion meditation (CM; Wallmark, Safarzadeh, Daukantaitė, & Maddux, 2013). A recent meta-analysis investigated the effectiveness of these interventions relative to control groups across 21 randomized control trials (RCTs) and identified significant between-group differences on self- Despite these beneficial effects on mental health and well-being, implementing compassion has revealed major limitations in some individual's abilities and motivations to develop compassion (Gilbert, 2010). Previous research has shown that some groups of individuals who might benefit most from cultivating compassion, also have major deficits in their abilities and motivation to cultivate compassion (Ebert, Edel, Gilbert, & Brüne, 2018;Gilbert et al., 2012;Gilbert, McEwan, Matos, & Rivis, 2011;Kelly, Carter, Zuroff, & Borairi, 2013;MacBeth & Gumley, 2012;Xavier, Gouveia, & Cunha, 2016). These groups include individuals experiencing a variety of traits, which include self-harm, self-criticism and shame, insecure attachment, alexithymia, low levels of empathy and mindfulness, increased symptoms of depression and anxiety, rumination and eating disorders. Early insecure attachment experiences, neglect, abuse, traumatization and excessive feelings of shame were identified as particularly relevant predictors for the development of fear of compassion for the self, for others and from others (e.g., Matos et al., 2017). These early affiliative experiences may lay down conditioned emotional memories in which the need for soothing, safeness and care becomes associated with fear, loneliness, sadness and grief (Gilbert, 2010;Liotti, 2004). In particular, traumatic experiences or memories of shame, which are of critical importance

Key Practitioner Message
• Fears of compassion have been found to impede progress in psychotherapy.
• The German version of the fears of compassion scale (FCS) is a reliable and valid measure to detect fears of compassion.
• The German version of the FCS and its subscales clearly discriminate between clinical and nonclinical participants.
• Patients with a diagnosis of borderline personality disorder show the strongest fears of compassion among the investigated clinical and nonclinical samples.
• The German version of the FCS is suitable to assess demands for specific psychotherapeutic interventions which can reduce fears of compassion.
for identity, may render one to feel inferior, defective, powerless and unattractive and to perceive others as critical, rejecting, condemning or abusive. These feelings will influence the formation of negative self-other schemas and engender a sense of ongoing threat to one's social self (Gilbert, 2010;Matos, Duarte, & Pinto-Gouveia, 2015;Matos & Pinto-Gouveia, 2014;Matos, Pinto-Gouveia, & Gilbert, 2013). Additionally, a lack of experience of security, safety and being nurtured as a child may lead to an undeveloped safeness-soothing system, which undermines one's ability to generate warmth and feel safe within social relationships and will also disrupt effective emotional regulation (Gilbert, 2009, Gilbert, 2010Matos & Pinto-Gouveia, 2014;Porter et al., 2020). Research supporting these assumptions indicates that a fear of compassion is predictive of lower oxytocin levels in patients with borderline personality disorder (BPD; Ebert et al., 2018). Consequently, the engagement in compassionate experiences or behaviours are linked to fears of being seen as weak or self-indulgent, of being judged or rejected due to compassionate efforts, of becoming too upset or overwhelmed by the needs of others when engaged in compassionate behaviours, and thus, the thinking that compassion will be viewed by others as manipulative or self-interested (Gilbert & Mascaro, 2017;Vitaliano, Zhang, & Scanlan, 2003). Furthermore, for individuals with high levels of self-criticism and interpersonal insecurity, being in compassion-based interventions may not produce soothing or safe effects, but rather increases stress, which can be measured using physiological indicators (Longe et al., 2010;Rockliff et al., 2008;Rockliff et al., 2011). In addition, strong fears of compassion have been shown to impede engagement, progress and outcome in psychotherapy Kelly et al., 2013;Merritt & Purdon, 2020).
To specifically examine resistance to compassion, Gilbert et al. for the fears of self-compassion subscale . The FCS is an internationally used instrument that has demonstrated promising validity and reliability across multiple studies (Cunha & Paiva, 2012;Gilbert, Clarke, Hempel, Miles, & Irons, 2004;Kupeli, Chilcot, Schmidt, Campbell, & Troop, 2013;Pinto-Gouveia, Castilho, Matos, & Xavier, 2013). A recent meta-analysis with data from 4,723 participants from clinical and nonclinical populations showed positive correlations between mental health difficulties (self-criticism, shame, depression, anxiety, distress and well-being) and fears of selfcompassion (r = .49), fears of compassion towards others (r = .30) and fears of compassion from others (r = .48). The strongest associations were found between the mental health variables of shame, selfcriticism and depression and the FCS subscales of fears of selfcompassion (FSC) and fears of compassion from others (FCFO). Overall, associations are significantly stronger for clinical populations than for nonclinical populations (Kirby, Day, & Sagar, 2019). Earlier findings have already demonstrated the FCS's potential to discriminate between clinical and nonclinical populations. The findings of a comparative study of 155 female undergraduate students and 97 females starting eating disorder treatment revealed significantly higher scores on the fears of self-compassion subscale in the latter sample (Kelly, Vimalakanthan, & Carter, 2014). A recent study compared the severity  (Dentale et al., 2017) and Japanese (Asano et al., 2017); the Italian and Japanese versions have already been validated (Asano et al., 2017;Dentale et al., 2017). A translation and psychometric evaluation of the FCS into German is missing. The purpose of this study was to provide a German translation of the FCS and to establish its psychometric properties, including internal consistency, as well as convergent and discriminant validity in a German sample.

| Diagnostic instruments
In the mixed clinical and BPD sample, the complete SCID-CV Wittchen et al., 1997) and SCID-II Fydrich et al., 1997) were conducted to determine the diagnostic status. The population-based sample received a screening of the SCID-CV and SCID-II, and they were asked whether they are in psychotherapeutic treatment and in which treatment setting (residential vs. outpatient) as part of the online study questionnaire.
The control group received the same screening as a telephone T A B L E 1 Steps in the translation process (1) The authors of the original version of the FCS were consulted for authorization. Three independent native German speakers who were fluent in English were determined.
(2) The original FCS was translated into German by the determined native German speakers.
(3) The three resulting translations were compared and merged into a single forward translation.
(4) The resulting German version of the FCS was translated back into English by an independent professional translator.
(5) The back-translation was reviewed by means of a comparison of the back-translated versions of the instrument and the original to highlight and investigate discrepancies between the original and the reconciled translation.
(6) To resolve discrepancies between back-translated versions of the instrument and the original, the items of the German version of the FCS were harmonized.
(7) The results were initially debriefed by testing the instrument on a small group of relevant people from clinical and nonclinical samples in order to test alternative wordings and check for the understandability, interpretation and cultural relevance of the translation.
(8) The test persons' interpretations of the translation with the original version were compared to highlight and amend discrepancies. Items were finalized.
(9) Items were reviewed a final time to highlight and correct any typographic, grammatical or other errors.
(10) A final report was written at the end of the process, documenting the development of each translation.
Note: The fears of compassion scales (FCS; Gilbert et al., 2011) was translated following the recommendations of the 'ISPOR Task Force for Translation and Cultural Adaptation' (Wild et al., 2005) using a 10-step procedure for translation.
interview. The presence of a current mental illness and current psychotherapeutic treatment were exclusion criteria for this sample.

| Self-criticism
The

| Data analytic plan
Internal consistencies of FCS total and subscale scores were assessed with Cronbach's α. Intercorrelations between subscales were calculated using Pearsons's correlation coefficient (one-tailed testing against 0). Similarly, convergent validity was assessed by computing Pearson correlations (one-tailed) between FCS subscale scores and scores on self-report measures of theoretically related constructs (i.e., SCS, RSES, TDEQ-12-SF, HADS, SWLS, AAS, SSASD). A Kruskal-Wallis test using χ 2 approximation was conducted to examine between group differences on age as the data were not normally distributed. Fisher's exact test was used to analyse between group differences on gender, education level, diagnosis and treatment setting.
Bonferroni corrected post hoc tests were computed for between group comparisons to avoid inflation of the type-I error. Due to significantly different variances in between group comparisons on the self-rating questionnaires, Sattherthwaite corrections were made.
Although the total FCS score was calculated in the data analysis of this study, it is recommended to always report the scores from the three subscales to differentiate the three dimensions of fears of compassion. Descriptives, standard psychometric analyses and internal consistency were conducted in SPSS 23.0 (IBM SPSS statistics).

| RESULTS
The majority of the n = 430 participants were female (78.3%); their mean age was 24.8 years (SD = 5.9, range = 18-72 years). Between group comparisons indicated significant differences concerning gender, age, education level and diagnosis. An overview of the four groups' characteristics including education and anamnestic data is provided in Table 2.

| Internal consistencies and intercorrelations between FCS total and subscale scores
Cronbach's α, means, SDs and intercorrelations of FCS total and subscale scores are displayed in Table 3. According to widely accepted standards (Cicchetti, 1994) internal consistencies for FCS total score and the three subscales were good to excellent in all four samples (Cronbach's α ranged from .76 to .96). As expected, all three subscales fears of compassion from others, fears of compassion towards others and fears of self-compassion were positively correlated with medium to large (r ≥ 0.3) intercorrelations in all four samples suggesting that the three dimensions are interrelated but not entirely overlapping. Internal consistencies of all remaining self-ratings are shown in Table 4.

| Between group differences on the FCS subscales
To examine the specificity of the FCS, that is, the potential to discrimi- For further between group comparisons for the FCS, see Table 5. Table 6 displays means and standard deviations and between group differences for the other investigated self-rated questionnaires.

| Convergent and discriminant validity of the FCS
The convergent validity of the FCS was investigated for the population-based and mixed clinical samples. As shown in Table 4  can possibly be attributed to qualitative differences between fear of compassion for others and fear as defined in the diagnosed mental disorders. Previous studies suggest that fear of compassion for others may be related to personality variables and empathy (e.g., Graziano, Habashi, Sheese, & Tobin, 2007), desired moral identity (e.g., Reed & Aquino, 2003), insecure attachment style (e.g., Feeney & Collins, 2001;Mikulincer, Shaver, Gillath, & Nitzberg, 2005), the personal significance of the recipient (e.g., Bakan, 2005), fear of being submissive, weak or being exploited by others (McLaughlin & Hughes, 2003).
From a methodological point of view, previous studies have found only small to medium correlations, even in significant larger samples than ours (e.g., Dentale et al., 2017;Kirby et al., 2019;Merritt & Purdon, 2020). In studies with small sample sizes, such as the Japanese translation and validation study (Asano et al., 2017) revealed that patients with BPD (a) were over 13 times more likely to report childhood adversity than nonclinical patients, (b) were more likely to report childhood adversity than other clinical populations, (c) reported elevated emotional abuse and neglect relative to controls (Porter et al., 2020). Several limitations should be considered when interpreting these results. First, there are differences between the samples in terms of gender, age and educational attainment. Previous meta-analytical findings on gender differences in self-compassion found that males reported slightly stronger fears of self-compassion than females (Yarnell et al., 2015). Nevertheless, in our study, demographic variables were found to have low, mostly nonsignificant correlations with FCS total and subscales. Second, the majority of individuals from our two patient samples were in residential treatment (82%), with 40% having two and 19% having three or more clinical diagnoses. Our results might, therefore, represent a specific population of individuals with relatively high levels of psychopathology and fears of compassion, and this should be considered when interpreting the very large between group differences found among FCS subscale scores. Third, due to the nonrepresentative distribution of specific mental disorders represented in our patient samples, group comparisons were only drawn between the predefined recruited samples, rather than comparing fears of compassion between mental disorders across samples. Despite these limitations, the results clearly indicate that the German version of the FCS exhibits satisfactory psychometric properties.
Fears of compassion are closely related to mental health difficulties (Kirby et al., 2019) and symptoms of psychopathology Gilbert et al., 2012;Kelly et al., 2013;MacBeth & Gumley, 2012;Xavier et al., 2016). These fears of compassion are also closely linked to poorer psychotherapeutic treatment outcomes (e.g., Kannan & Levitt, 2013;Marshall, Zuroff, McBride, & Bagby, 2008;Rector, Bagby, Segal, Joffe, & Levitt, 2000). Future research should test the FCS's potential to assess manifestations of psychopathology and demands for specific psychotherapeutic interventions, for example, CFT (Gilbert & Procter, 2006), across different clinical groups. Because our results indicate the BPD sample reported the strongest fears of self-compassion and fears of compassion from others, it should be further determined whether this feature is particularly pronounced in individuals with a diagnosis of BPD, who might then require more intensive treatment.
In conclusion, the current study suggests that the German version of the FCS is a reliable and valid instrument for measuring fears of compassion. Furthermore, the German FCS exhibits sufficient specificity to assess mild to severe manifestations of fears of compassion and the ability to differentiate between individuals from the general population and between clinical contexts. Thus, the German FCS is a promising instrument for detecting potential obstacles to psychotherapeutic treatment progress. Finally, due to its close relations to physical and psychological health as well as life satisfaction and social relationships, the German FCS can be used as a useful measure of treatment outcomes.

ACKNOWLEDGEMENT
Last authors contributed equally to the study. Open access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST
No conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are currently available on request from the corresponding author. An active link containing repository name, URL and reference number will be included in the final accepted manuscript.