This was a retrospective cohort study. No ethical approval was required for this secondary analysis of existing data from previous medical records. Informed consent was obtained from all participants before treatment inclusion, and the study was approved by and conducted in accordance with the hospital's ethics review board. The eligibility criteria for the current study were patients who met the DSM-5 diagnostic criteria for SSD with more than 6 months' duration, who were aged between 18 and 65 years, and who underwent an 8-week MBCT program in Jining No.1 People's Hospital between July 2018 and December 2020. Over 400 patients with SSD were recruited through outpatient clinic screening and allocated to receive either a psychiatric consultation intervention or MBCT (Figure 1).

Participants were excluded from receiving the MBCT intervention if they had severe psychopathology (psychotic disorder, bipolar disorder, depressive disorder, substance-related disorders, severe personality disorders, etc.), a severe and unstable physical illness, cancer diagnosed within 5 years, a history of CBT, significant cognitive or visual impairment, or poor language skills that would hinder comprehension of the intervention or measurement. Patients who underwent fewer than 4 sessions of MBCT or for whom data on evaluated parameters were missing were also excluded from this study. Data and patient identification details were anonymized before analysis.

The MBCT intervention was conducted by one licensed mental health specialist (JS) with more than 2 years of experience in independently delivering MBCT education and training to ~100 patients. The MBCT program used in this study was based on the MBCT format for patients with depression (34). The Chinese MBCT program has been successfully applied in other studies (35, 36). We made minor modifications to render the MBCT training protocol suitable for the clinical setting of SSD. First, the cognitive components of MBCT included psychoeducation in respecting physical and mental boundaries in session 3 and identifying negative automatic thoughts in session 4. Second, to make it easier for patients to participate in the MBCT program, sessions were scheduled once a week or every other week for 8 consecutive weeks, with sessions lasting 90 min, in groups of 3–5 participants; the daily homework exercises guided by audio recordings after the weekly sessions were shortened to 30 min per day; and the content of the precourse introduction was integrated into session 1. Third, loving-kindness meditation was introduced in session 2 to cultivate kindness within participants and compassion for their somatic symptoms. Practices in the MBCT program mainly included guided body scan, mindful breathing, mindful walking, mindful awareness, loving-kindness meditation, and sitting meditation. Medical education on how to use medications properly also was added into the MBCT program. In each session, the therapist introduced the therapeutic modules and explained how to carry out the mindfulness exercises. Adequate participation in the MBCT program was defined as having attended at least 4 sessions, as in previous trials (37). The study data were collected upon admission to the MBCT program and at 6-month follow-up following treatment inclusion using self-report questionnaires or measures administered via face-to-face interviews.

All data, including demographics and evaluated variables, were obtained by reviewing electronic medical records. Demographic data included age, gender, illness duration, whether the patient was taking prescription antidepressants, marital status, education level, and employment status. Evaluated variables included clinical and process outcomes. The clinical outcome measure was the GIC score at 6-month follow-up following treatment inclusion (post-treatment follow-up). The process outcome measures were BSI, SCS, and TAS scores at treatment inclusion (pre-treatment) and post-treatment follow-up.

Data analysis was conducted using SPSS (version 26.0; SPSS Inc.). The tested mediation model was estimated using MEMORE (version 2.1) for SPSS. Continuous data were tested for normality using the Shapiro–Wilk test, and if normally distributed, the data were expressed in the form mean ± standard deviation (SD) and tested using paired t-tests for intergroup comparisons. Changes in measures from pre-treatment to post-treatment follow-up were summarized with means (95% confidence intervals). To assess the differences in post-treatment follow-up measures between patients taking antidepressants and those not taking antidepressants, linear models were used with the GIC score at post-treatment follow-up as the outcome variable and the indicator of antidepressants as the independent variable, adjusting for other pre-treatment covariates. Cohen's d effect sizes were calculated by dividing the mean difference between pre-treatment and post-treatment follow-up by the standard deviation of the difference; Cohen's d values of 0.20 to 0.50, 0.50 to 0.80, and > 0.80 represent small, moderate, and large effects, respectively (48). Bivariate correlations were calculated to examine the relationships between changes in psychometric variables of the Brief Symptom Inventory General Symptom Index (BSI-GSI), SCS (SCS-pos, SCS-neg, and total score), and TAS (DIF, DDF, EOT, and total score); correlation coefficients (r) of < 0.10, 0.10 to 0.30, 0.30 to 0.50, and 0.50 to 1.00 were considered to represent weak, small, moderate, and strong correlations, respectively (48, 49). Multivariate linear regression analysis was performed to assess the potential mediation effects of changes in the SCS dimensions (SCS-pos and SCS-neg) and the TAS dimensions (DIF, DDF, and EOT) as predictors of the GIC score and change in the BSI-GSI with adjustment for demographics and the corresponding pretreatment score. A within-subjects serial multiple-mediator model was further used to assess the mediating effects of self-compassion and alexithymia on the BSI-GSI. Confidence intervals for parameter estimates were determined from 5,000 bootstrapped samples. The threshold for statistical significance was p < 0.05.

One hundred fifty-two patients were enrolled in an 8-week MBCT program, among whom 19 subjects attended fewer than 4 treatment sessions, 12 subjects were lost to follow-up at 6 months following treatment inclusion, and measurement data were missing for five subjects. Thus, 116 participants were eligible for this cohort analysis, of whom 32 (27.6%) were taking prescribed antidepressants upon entering the program (Figure 1). The demographics and clinical characteristics of the 116 patients who met the analysis criteria are presented in Table 1. The clinical improvement rate following MBCT was 64.7% as indicated by GIC scores. No difference in clinical improvement was found between patients taking prescribed antidepressants and those who were not upon entering the program.

The psychometric outcomes from pre-treatment to post-treatment follow-up are shown in Table 2. Compared with the pre-treatment values, the post-treatment values of the total score and the dimensional scores on the BSI-18, the SCS, and the TAS decreased or increased significantly (all p values < 0.001).

The relationships among the changes in psychometric variables are presented in the correlation matrix in Table 3, indicating that most of the correlations were significant and in the expected direction.

Standard linear regression was conducted to examine the potential mediating effects of improvement in self-compassion and alexithymia on the clinical outcome and psychosomatic distress. As shown in Table 4, decreases in SCS-neg, DIF, and EOT values and increases in SCS-pos values were each associated with GIC score and with a greater reduction in BSI-GSI value when adjusting for pretreatment covariates.

We further sought to explore how concurrent changes in SCS values and TAS values as mediators impacted perceived psychosomatic distress, as indicated by BSI-GSI scores, using a serial multiple-mediator model. In serial mediation, each mediator is assumed to causally influence subsequent mediators: for instance, an increase in self-compassion is predicted to cause a decrease in alexithymia. Given that these are causal inferences, the estimated effects are unidirectional (MBCT→ΔSCS→ΔTAS→ΔBSI-GSI). The present serial mediation model revealed that changes in both SCS values and TAS values had significant indirect effects on the pre–post change in BSI-GSI value [ΔSCS β = −1.810 (a 1*b₁), bootstrap SE = 0.341, 95% bootstrap CI (−2.488, −1.160); ΔTAS β = −1.615 (a 2*b₂), bootstrap SE = 0.391, bootstrap 95% CI (−2.413, −0.896); ΔSCS→ΔTAS β = −0.621 (a 1*a 3*b₂), bootstrap SE = 0.183, bootstrap CI (−1.032, −0.315)], as shown in Figure 2; see Supplementary Table S1 for the full model [F_(4,111) = 31.464, p < 0.001, R² = 0.531, MSE = 5.924].

In contrast, post-hoc analysis with a reverse sequence (MBCT→ΔTAS→ΔSCS→ΔBSI-GSI) revealed that only the change in TAS value still had a significant indirect effect on pre–post change in BSI-GSI value [ΔTAS β = −2.235 (a₁'^*b₁'), bootstrap SE = 0.516, bootstrap 95% CI (−3.305, −1.270); ΔSCS β = 0.013 (a'2*b₂'), bootstrap SE = 0.323, 95% bootstrap CI (−0.600, 0.682); ΔTAS→ΔSCS β = −1.823 (a₁'^*a₃'^*b₂'), bootstrap SE = 0.439, bootstrap CI (−2.770, −1.047)], as shown in Figure 3; see Supplementary Table S1 for the full model [F_(4,111) = 31.464, p < 0.001, R² = 0.531, MSE = 5.924].

Several limitations should be considered when interpreting the results of our analysis. First, this was a single-center retrospective study with a small sample size. There was some degree of selection bias. Second, no participant was recruited into a control group in this study, which limits the ability to compare the outcomes with natural changes in psychosomatic distress, self-compassion, and alexithymia over time. Third, this study was only designed to explore the impact of implicit self-compassion on alexithymia. The MBCT program is a multi-component program. Other active components that were not assessed in this study, such as mindfulness, acceptance, emotions, cognition, and behaviors, need to be tested for association with alexithymia. Previous research has suggested that increased mindfulness is associated with a decrease in DIF (60). Additionally, many participants were also treated with antidepressants, and the types of antidepressants varied. Changes in self-compassion and alexithymia following the MBCT intervention might be enhanced by these concomitant medications. Fourth, alexithymia as a personality trait is transdiagnostic and linked to a range of psychiatric morbidities, such as depression, panic disorder, eating disorders, chronic pain disorders, substance abuse, and suicide (53). The findings from a small sample of SSD patients were limited in their ability to guide clinical practice in terms of generalizability.