The intervention lasts for 3 months and includes 12 sessions that follow a group format of 2 hours and with a weekly frequency. It includes the following components: Psychoeducation about personality disorders and, specifically, BPD and how these disorders evolve. Problems associated with BPD (eg, alcohol and / or drug use, eating disorders, etc.). The importance of "modeling" in family members: change of attitude towards difficult situations to reduce the escalation of tension and re-establish a healthy relationship between the patient and family. The handling of problems. Crisis management: Development of safe plans for when the patient is in emotional escalation.

We used the spanish version of the instrument Difficulties in Emotion Regulation Scale (DERS), developed by Gratz y Roemer (2004), with the aim to assess the emotional regulation problem. The scale was adapted to spanish and it was reduced from 36 items to 28 items (Hervás & Jódar, 2008). In this version, they figured on 5 scales instead 6 considering that "difficulties in the impulse control" and "limited access to regulation strategies" were joined into one named "emotional lack of control". The subscales are life interference, emotional confusion, emotional rejection, emotional lack of control and lack of emotional attention. Items are rated on a 5-point Likert scale ranging from 1 (rarely) to 5 (almost always), and higher scores indicate greater impairment or dysregulation. The psychometric properties show high consistency were Cronbach's alphas of the subscales range from .73 to .91 and .93 for the total scale (Hervás & Jódar, 2008).

Family Empowerment Scale (FES) consists of 34 items divided in three subscales: family, service system, and involvement in community that is refered to three ways of empowerment, attitudes, knowledge, and behaviors (Koren, DeChillo & Friesen, 1992). Items are rated on a scale of 1 (completely false) to 5 (totally true), and higher scores indicate a greater sense of empowerment. The psychometric properties are the following: regarding to the internal consistency of FES subscores, the coefficients ranged from .87 to .88 and validity and reliability are adequated (Koren, DeChillo & Friesen, 1992).

Connor-Davidson Resilience scale is a 25-item measure of resilience. Items are rated on a 5-point Likert scale ranging from 0 (absolutely not) to 4 (almost always) and the punctuation is based on how the participant has felt over the last month. Higher scores means greater resilience (Connor & Davidson, 2003). The CD-RISC authors reported acceptable test-retest reliability (r = 0.87) and strong internal consistency (α = .89) (Connor & Davidson, 2003).

Quality of Life Index-Spanish Version consist of 10 items that assess aspects as physical, psychological/emotional well-being, self-care and independent functioning, occupational and interpersonal functioning, social-emotional and community and services support, personal and spiritual fulfillment and global perception of quality life. Items are rated on a scale of 1 (bad) to 10 (excellent), and higher scores means higher quality of life. This instrument has good psychometric properties as Cronbach's alpha of .89 and high test-retest reliability (r = 0.87) (Mezzich et al., 2000).

Beck Hopelessness Scale is a 20-item true-false instrument utilized for the measurement of hopelessness. It evaluates the attitude of the participant in the previous week. Nine items assess attitudes about the future and eleven items assess pessimistic statements. Items are rated as true or false, and higher scores reflects higher levels of hopelessness. Good psychometric properties are shown in this instrument. Internal consistency was excellent (α = .93) (Beck, Weissman, Lester & Trexler, 1974).

Openness Towards the Future Scale is a 10-items self-report that measures positive affective orientation towards the future. Items are rated on a 5-point Likert scale ranging from 1 (totally disagree) to 5 (totally agree), and higher scores indicate better openness to the future. It shows adequate psychometric properties for both clinical and general samples. Cronbach's alpha was acceptable for both clinical (α = .82) and community samples (α = .87) (Botella et al., 2018).

Personality Inventory for DSM-5 (PID-5) is a 220-item instrument that assesses 25 traits, which form five domains (Negative Affectivity, Detachment, Antagonism, Disinhibition, and Psychoticism). The 25 traits are the following: Anhedonia, Anxiousness, Attention Seeking, Callousness, Deceitfulness, Depressivity, Distractibility, Eccentricity, Emotional Lability, Grandiosity, Hostility, Impulsivity, Intimacy Avoidance, Irresponsibility, Manipulativeness, Perceptual Dysregulation, Perseveration, Restricted Affectivity, Rigid Perfectionism, Risk Taking, Separation Insecurity, Submissiveness, Suspiciousness, Unusual Beliefs and Experiences, and Withdrawal. Items are ranged on a 4-point Likert scale ranging from 0 (totally disagree) to 3 (totally agree), and higher average scores indicate more "dysfunction" in a specific personality trait facet or domain. Internal consistency ranged from .72 to .96 with a median of .86 (Krueger, Derringer, Markon, Watson & Skodol, 2012).

Structured Clinical Interview for DSM-IV Axis II consists of a semistructured format involving 10 standard DSM-IV personality disorders and the Personality Disorder Not Otherwise Specified, Depressive Personality Disorder and Passive-Aggressive Personality Disorder. Items are rated as "yes" or "no", and two or more scores on "yes" mean that participant may have a personality disorder and should be evaluated with the interview, and you have to assess the personality disorder depending on the punctuation of each (5 items or more: dependent, depressive, squizotypal, histrionic, narcissistic and borderline; 4 items or more: avoidant, obsessive-compulssive, passive-agressive, paranoid and schizoid; 3 items or more: antisocial; 1 item or more: non-specified. Adequate inter-rater reliability coefficients were reported (.48-.98) and satisfactory internal consistency coefficients (ranges from .71 to .94) (Maffei et al. 1997).

This instrument is also carried out in the evaluation of patients. Depression, Anxiety and Stress Scale (DASS) have 42 items about negative emotional symptoms (Lovibond & Lovibond, 1995). Lovibond & Lovinbod (1995) proposed that a part of these subscales for can become part of a short version creating a new questionnaire of 21 items. Items are rated on a 4-point Likert scale ranging from 0 (It did not happen to me) to 3 (It happened to me a lot or most of the time), and higher scores indicate worse symptoms of depression, anxiety or stress. DASS-21 showed fantastic factor structures. Regarding to the internal consistency, Cronbach's alphas were excellent for the DASS-21 subscales: Depression (α = .94), Anxiety (α = .87) and Stress (α = .91) (Antony, Bieling, Cox, Enns & Swinson, 1998).

This instrument is also carried out in the evaluation of patients. Difficulties in Emotion Regulation Scale (DERS), developed by Gratz y Roemer (2004), with the aim to assess the emotional regulation problem. The scale was adapted to spanish and it was reduced from 36 items to 28 items (Hervás & Jódar, 2008). In this version, they figured on 5 scales instead 6 considering that "difficulties in the impulse control" and "limited access to regulation strategies" were joined into one named "emotional lack of control". The subscales are life interference, emotional confusion, emotional rejection, emotional lack of control and lack of emotional attention. Items are rated on a 5-point Likert scale ranging from 1 (rarely) to 5 (almost always), and higher scores indicate greater impairment or dysregulation. The psychometric properties show high consistency were Cronbach's alphas of the subscales range from .73 to .91 and .93 for the total scale (Hervás & Jódar, 2008).

Lum Emotional Availability of Parents consists of 15 items that involves participants' mothers and fathers emotional availability. Items are rated on a 6-point Likert scale ranging from 1 (never) to 6 (always), and higher scores indicate better emotional availability. Internal consistency was excellent for non-clinical sample for the mother form (α = .96) and for father form (α = .97); also for the mother form in a clinical sample (α = .92), and for the father form (α = .93). This instrument has adequated test-retest reliability for the mother form (r = .92) and for the father form (r = .85) (Lum & Phares, 2005).

Validating and Invalidating Responses Scale is a 16-items self-report which evaluates levels of validation and invalidation of caregiver's responses. This instrument has two subscales: validation and invalidation responses. This two are moderately correlated. Items are rated on a 5-point Likert scale, ranging from 0 (never) to 4 (almost all the time), and higher scores indicate more perceived validation or invalidation from the caregiver who is assessed. There are no psychometric properties available on the VIRS yet.

Post-Module is an instrument constructed by our research team oriented to assess the level of change obtained with respect to the therapeutic modules as well as the degree of satisfaction to receive the treatment. It evaluates the six modules of the treatment in the two conditions. There are two subscales: one evaluates the learning of the abilities of the module and it is rated from 0 (nothing) to 10 (a lot), and the other evaluates how the module has helped to the caregiver to improve several aspects such as knowing and understanding the problem, emotion's comprehension, mindfulness of the relationship with their relative, acceptance, family atmosphere and problem solving on the family atmosphere, and it is rated from 1 (nothing) to 4 (a lot). Expectation scale is assessed on the end of the first module.