Elevated HRV, which refers to greater variability and higher vagal tone, signifies greater stress adaptability (12). Additionally, it has been associated with more positive emotions, more enjoyment of social interaction, less negative emotions during stressful tasks and higher social connectedness (13–15). In contrast, lowered HRV has been suggested to be a marker for all types of diseases and can predict mortality (16, 17). In correspondence to this, various mental diseases such as panic disorder, posttraumatic stress disorder and depression are linked with an imbalance in the ANS, as evidenced by attenuated HRV compared to healthy controls (18–20). The Neurovisceral Integration Model, introduced by Thayer and Lane (21), offers a possible explanation regarding the relationship between psychopathology, stress, and HRV. It postulates that stress could impact brain function, thus hampering adaptive and flexible behavior as indicated by low HRV (10, 22, 23). Therefore, developing effective treatment programs to counter stress associated with psychopathology is of special interest in scientific research (24).

Treatment methods of high relevance and increased recognition in recent years are MBI (25, 26). These techniques are about accepting the reality we live in, letting go off thoughts and living in the here and now, which could facilitate psychological well-being (27). Evidence showed efficacy of MBI in reducing stress, anxiety, and depression in clinical populations (28–30) as well as in healthy subjects (31–33). Furthermore, they appeared to be linked with a substantial increase in self-compassion, quality of life, coping with problems, happiness, resilience and overall psychological well-being (34–37).

A specific MBI essential in the present study is the MSC program founded by Neff and Germer (38). Self-compassion has gained increased attention since its introduction in 2003 (39) and can be described as “the ability to be kind and understanding toward ourselves when we suffer, fail, or feel inadequate” (40, p. 861). The program contains the three important aspects: self-kindness versus self-judgment, mindfulness versus over-identification, and common humanity versus isolation (40). MSC comprises an 8-week program, with one session each week (2.5 h) and a half-day silent meditation retreat (38). It consists of guided sessions including various topics regarding self-compassion (i.e., practicing loving-kindness) and additional homework which enables to practice self-compassion in a formal (sitting meditation) as well as an informal (during everyday life) way (38, 40). Neff and Germer (38) demonstrated that the MSC program improved self-compassion, mindfulness, sympathy for others, life happiness, as well as well-being, and lowers stress, anxiety, and depression.

Self-compassion has shown to be repeatedly associated with well-being and coping with unpleasant or stressful life events (41–46). People high in self-compassion showed decreased worry, rumination, subjective stress, and positive emotion regulation (47–49). In sum, self-compassion is a well-established program to reduce psychological stress and help to better cope with stressful situations (50–52). However, to the authors’ knowledge, there are few findings on self-compassion and physiological stress reactivity. In general, research demonstrated that mindfulness could buffer physiological stress responses (53, 54). Correspondingly, literature showed an increase in HRV and decrease in BP during mindfulness training (55–57). Additionally, self-compassion and mindfulness training could lead to better stress coping and buffer physiological stress (58–61). However, findings are heterogenous (62–64).

In the psychiatric rehabilitation clinic Sonnenpark Neusiedlersee in Rust, Austria, the MSC program was modified to fit the typical 6-week length of stay at the clinic to see if this form of MSC shows positive results. While some of the meditations and exercises, such as the inquiry, are identical to the 8-week MSC course, other parts are modified to suit the patients need in the rehabilitation clinic. Gaiswinkler et al. (51) demonstrated higher self-compassion and happiness after 6-week MSC program in comparison to the active control intervention of Progressive Muscle Relaxation (PMR) in this rehabilitation clinic. Moreover, they observed psychiatric and quality of life parameters improving in both groups to the same extent (51). PMR is a very well-established and empirically strong validated relaxation technique (65). Individual muscle groups are tensed and loosened immediately after to increase inner relaxation (66). More specifically it can be useful to decrease stress (67). Since it does not include the mindful/ self-compassion aspect it previously has been chosen as an adequate option for an active control group (68, 69).

The primary aim of this study was to investigate if rehabilitation in-patients who received MSC training show less pronounced stress reactions compared to patients who received PMR training regarding psychophysiological stress reactivity using a social-evaluative stress paradigm. HRV, HR, BP, Positive and Negative Affect (PA/NA), and Subjective Stress Perception (SSP) were compared between groups. So far, the MSC program has only been explored on a psychometric level (51). Therefore, this study served to explore the MSC program on a biometric level in a clinical setting. We expected lower physiological responses to the stressor in the MSC group (i.e., lower HRV decrease, HR increase, and blood pressure increase, from baseline to stress). For secondary outcome, we intended to investigate whether Self-Compassion, Emotion Regulation (Cognitive Reappraisal and Suppression), and Appraisal are higher in the MSC compared to the PMR group. Furthermore, we expected that Rumination and Psychological Distress should be higher in the PMR group.

The presented study is part of a broader study (n = 170) which has not yet been published. Within that study, patients were randomly assigned into two groups (MSC, PMR) prior to their start of the rehabilitation stay. For the present study, the randomized allocation was adopted. Participants were adults attending a 6-week psychiatric rehabilitation stay at Sonnenpark Neusiedlersee clinic, located in Burgenland, Austria. All exclusion criteria of the rehabilitation clinic applied to the current study, such as acute suicidal and psychotic episodes or acute addiction disorders, determined by the treating psychiatrist at the clinic. Besides, people with severe cardiovascular diseases were not included. In this investigation, 59 participants were recruited of which nine were excluded (4 = drop out, 3 = no speech, 1 = cardiac arrythmia, 1 = previous stroke). All participants signed a consent form and the investigator ensured that participation was voluntary and withdrawal from consent throughout possible. After the examination, they got a coffee voucher from the in-house café. A sample of 50 individuals (35 female) aged from 19 to 76 (M = 47.22, SD = 12.44) was examined (25 MSC, 25 PMR). The study was ethically approved by the ethics committee of the University of Graz (GZ. 39/84/63 ex 2020/21).

Participants were tested in the last week of their stay at the clinic (post-test only experimental design; see Figure 1). They were requested not to consume caffeine, alcohol, nicotine, and sugary drinks, as well as to refrain from physical exertion 2 h before the testing (73, 97, 98). Initially, participants’ height, hip, and waist circumference were measured, and anamnestic data collected. During the recording, participants sat quietly in an upright position and were asked not to move (98). HRV and HR were measured continuously. BP was measured 6 times (baseline & recovery: min 2:30, 4:00; speech prep.: min 0:30, 2:00). For the 5-min baseline recording, landscape pictures were presented on a screen. After the 3-min speech preparation, the actual 3-min speech took place. The recovery phase followed, identical to baseline. After baseline, speech preparation, and recovery PANAS and SSP were presented. Appraisal was collected after speech preparation and recovery and Rumination after recovery. Finally, participants were debriefed. It was made clear that the video recordings were not evaluated at all and were deleted immediately, but only served to reinforce stress (99). One female judge (S.K.) was present during all stress induction, wearing a white coat. There was always the same judge (S.K.). One camera was mounted right beside the computer screen. SCS-D, ERQ, and BSI-18 were presented on the last day of the 6-week stay. The time of day when the stress induction took place could not be controlled, otherwise it would not have been possible to give all patients an appointment for testing.

HRV analyzes were conducted with the software packages AcqKnowledge 4.3 (100) and Kubios HRV Premium 3.3.1 (101). R-R time series (interbeat intervals) were interpolated with 4 Hz (101). Fast Fourier Transformation (FFT) was applied to quantify HF-HRV [0.15–0.4 Hz] and LF-HRV [0.04–0.15 Hz] with 180 s window and no overlap (73, 102). The smoothness priors algorithm was used for detrending (101). To control for artifacts, HRV data were visually inspected by the examiner and corrected with the automatic correction algorithm in Kubios software if necessary (101). For further analyzes, only data containing less than 5% of artifacts were considered. An exception was made for one subject due to the small violation of the limit during recovery (artifacts = 5.16%). One person had to be excluded due to excessive artifacts (baseline = 23.63%, speech prep. = 20.09%, recovery = 25.85%). In addition, one participant could not be included in HRV, and HR analyzes due to flawed recordings. RMSSD, HF, SDNN, LF, and HR means of the last 3 min of the 5 min recordings (baseline, recovery) as well as means of the 3 min recording (speech prep.) were analyzed which seemed a sufficient length for ultra-short term HRV measures (74). Due to the sensitivity to movement in HRV recordings, an a priori decision was made to incorporate speech preparation as a stress phase (73). Research shows, stress anticipation can also trigger adequate stress responses (103). Prior to analyses, RMSSD and SDNN data were subjected to a natural log transformation to account for skewness (73, 104).

Statistical analyzes were done with SPSS, version 27 by IBM. Group comparisons regarding anamnestic variables were conducted by means of chi-square and unpaired t-tests. For primary outcome, separate two-way mixed ANOVAs with the between-subject factor group (MSC, PMR) and the within-subject factor time (baseline, speech prep., recovery) regarding psychophysiological measures (HRV, HR, BP, PA, NA, SSP) were performed. For post-hoc analyzes, Bonferroni pairwise comparisons were implemented. Secondary outcome analyzes included unpaired t-tests with the independent variable group (MSC, PMR) regarding Self-Compassion, Emotion Regulation, Psychological Distress, Appraisal, and Rumination. The statistical significance was set to p < 0.05 (two-sided). Alpha-error-accumulation was controlled via Bonferroni correction.

Participants were on average overweight [BMI: M = 28.01, SD = 5.93; (105)]. All individuals had a psychiatric diagnosis and 33 (66%) of them at least one comorbid psychiatric diagnosis as reported by the International Classification of Diseases [ICD-10; (106), Chapter F), presented in Table 1. Primarily, people with affective disorders (F30-F39) participated in the study [44 (88%) people; (106)], followed by neurotic, stress-related, and somatoform disorders [F40-F49; (106)]. Thirty-eight (76%) participants took psychotropic drugs, all of them antidepressants, followed by additional neuroleptics intake (14 (36.8%) people). Of note, 24 (48%) participants reported doing sports on a regular basis. Sixteen (32%) people smoked regularly, and four (8%) participants drank alcohol habitually.

Group comparisons regarding anamnestic data (health-related variables, psychiatric diagnoses, medication) did not reveal significant differences (see Table 1).

Forty-eight participants were included in the HRV and HR-analyzes, 47 for Respiration and 44 for the BP analyzes. If there were no more than 10% missing data in a questionnaire, the mean of the remaining items was inserted for the missing item (107). This procedure was applied for three participants for PA, two for NA and three for SCS-D. For SSP, Appraisal, and Rumination 49 people were enclosed. Data of 44 participants were included in the analyses of SCS-D and BSI-18. For ERQ Reappraisal, 42 people and for ERQ Suppression, 43 people were included (see Tables 1, 2).

Published resting short-term HRV norms of approximately 5 min length collected in 21,438 non-clinical individuals in sitting or lupine position (74) were compared to the 3 min baseline measurement in this clinical sample for exploratory reasons. The present sample showed significantly lower RMSSD, HF-HRV, and SDNN during baseline measurement as compared to the non-clinical norms. For LF-HRV, no significant difference was found (see Table 3). According to Shaffer and Ginsberg (74) participants should breathe at 11–20 breaths per minute (BPM) that short-term HRV values are adequate. Respiration rate for baseline measurement was on average 15 BPM (M = 15.30, SD = 4.46), for speech preparation 17 BPM (M = 17.21, SD = 3.12), and for recovery 15 BPM (M = 15.41, SD = 4.86). Mixed ANOVA with Respiration rate showed no significant main difference between MSC and PMR group (see Table 2). Therefore, no further analyzes including Respiration deemed necessary.

For all Mixed ANOVA calculations, prerequisites have been checked. When the assumption of sphericity was not fulfilled, the Greenhouse–Geisser correction was applied (see Table 2^a) (109). Normal distribution was assessed by Shapiro–Wilk test (p > 0.05), and if violated, ANOVAs were still calculated due to the central limit theorem (110). Box’s Test showed homogeneity of covariances for all calculations [p > 0.05; (111)]. Except for SSP in recovery phase, homogeneity of variances was given by Levene’s test (p > 0.05). Post-hoc comparisons were still interpreted for SSP (112).

For unpaired t-tests, normal distribution was given, as assessed by Shapiro–Wilk test (p > 0.05) except for PMR group regarding Psychological Distress and Rumination (110). Homoscedasticity was evident in all calculations (Levene’s test: p > 0.05). For Self-Compassion, Suppression, Psychological Distress, Appraisal, and Rumination, no group differences were detected. The MSC group showed significantly higher values in Reappraisal than the PMR group (see Table 1), indicating a medium effect size (Cohen’s d = 0.64; ns after Bonferroni correction). However, this result was non-significant after Bonferroni correction.

The first point to be mentioned is the sample size in this study. This only allows for large effects to be revealed by statistical evaluation (136, 137). To detect medium and small effects, a larger sample size is certainly required. Additionally, a within subject design study would be desirable (73). Quintana and Heathers (131) recommend collecting multiple data from one individual at several times in psychophysiological studies. Due to the selected stressor, we were not able to apply the speaking task twice, otherwise a learning effect and therefore an inadequate stress response might have been detected (73). For future studies, however, we should consider presenting a different stress task at the beginning of the 6-week stay as well as at the end to reveal the changes within a person pre- and post-treatment (73). Recently, Asbrand et al. (138) showed that the standardized TSST could be used in repeated measures, which might constitute an additional approach for future replication studies.

Besides, a non-clinical control group additionally to the active PMR control group would be essential to compare clinical and non-clinical subjects (139). In this study, an attempt was made to compare HRV norm values in non-clinical studies with the present sample, which indicated lower HRV values in comparison to norms. However, to our knowledge, no norm values exist for stress reactivity (74). Still, studies show blunted HRV reactivity in a stress experiment in patients with mental illness as compared to healthy individuals, which is in line with the Neurovisceral Integration Model and makes comparative values preferable in future studies (18, 19, 21, 79, 140). Considering that affective disorders are most frequent in the rehabilitation clinic, this is of particular interest since Jandackova et al. (141) consider HRV as an influencing factor on the onset of depression.

In the present study, the two instructors, teaching MSC and PMR, were not supervised. Thus, it would be interesting to address the training of the instructors as well as their mindset. Showing compassion for others requires us to be aware of our own pain and may help clinicians be more effective in therapy (142).