From September 2020 to January 2021, we retrospectively analyzed clinical data from 37 obese subjects and 37 demographic well-matched healthy people. All participants (1) aged above 18 years old; (2) could read and understand the description of each item of the questionnaire; (3) be voluntary participation in the survey and had signed the informed consent. The ethics committee had approved the study at Shanghai Sixth People's Hospital. All procedures followed the Declaration of Helsinki.
Basic demographic information (age, education years, height, weight) and a series of clinical scales (Childhood Trauma Questionnaire, CTQ; Beck Anxiety Inventory, BAI; Beck Depression Inventory, BDI and Eating Disorders Inventory, EDI) were collected. Body Mass Index (BMI, equals weight (kilogram) divided by height (metre) squared) was calculated to describe the severity of obesity.
Childhood Trauma Questionnaire (CTQ) is designed for adolescents and adults to obtain a brief, reliable and valid assessment of traumatic experiences in childhood (Bernstein et al., 1994; Bernstein et al., 2003). It assesses the incidents of abuse and neglect in childhood, including physical abuse, emotional abuse, sexual abuse, emotional neglect, and physical neglect (Bernstein, Ahluvalia, Pogge, & Handelsman, 1997). CTQ has 28 items, and each item adopts a 5-point Likert score from 1 "never " to 5 "very often" according to the frequency of the experiences that occurred (Bernstein et al., 1994). A higher CTQ score indicates more severe childhood trauma. The total Cronbach's α of the Chinese version CTQ is 0.73 (MD Zhang, 2011).
Beck Anxiety Inventory (BAI) assess the severity of generalized anxiety symptoms (A. T. Beck, Epstein, Brown, & Steer, 1988). It had good reliability, validity, internal consistency, and convergence (Fydrich, Dowdall, & Chambless, 1992; Muntingh et al., 2011). The total Cronbach's α of the Chinese version BAI is 0.95 (Kin-Wing, 2002). BAI has 21 items, with each response based on a 4-point Likert scale ranging from 0 means "not at all" to 3 means "severely". A higher score indicates greater anxiety severity.
Beck Depression Inventory (BDI) version 2 is a widely used clinical instrument to evaluate the depression severity in normal populations (Aaron T. Beck, Steer, & Carbin, 1988; Kühner, Bürger, Keller, & Hautzinger, 2007; Whisman, Perez, & Ramel, 2000). It had good reliability and validity (Kühner et al., 2007). The total Cronbach's α of the Chinese version BDI is 0.94 (Lu, 2011). BDI has 21 items, and each item consists of four self-evaluative statements scored 0 to 3, with an increasing score indicating greater depression severity.
Eating Disorder Inventory version 2 (EDI) measures the eating disorder symptoms and the cognitive and behavioral characteristics of anorexia nervosa, bulimia. EDI contains 91 items, with each response based on a 6-point Likert scale ranging from 0 "Never" to 5 "Always" (Garner, Olmstead, & Polivy, 1983). It consists of 11 subscales. Three measuring the primary eating disorder symptoms: (1) Drive for thinness; (2) Bulimia; (3) Body dissatisfaction. Eight measuring correlated psychological traits: (4) Ineffectiveness; (5) Perfectionism; (6) Interpersonal distrust; (7) Interceptive awareness; (8) Maturity fear; (9) Asceticism; (10) Impulse regulation; (11) Social insecurity.
To test the common method biases, we performed Harman's single-factor test to exam this issue (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Unrotated factors analysis revealed that the eigenvalues of 27 factors were >1. The first factor explained only 26.42% of the variance, which was much lower than the threshold of 40%. The result indicated the absence of severe common method biases in the investigation.
Statistical analysis was carried out with SPSS 21. The comparison between obesity and healthy control was performed through the independent t-test (sex distribution comparison was verified through χ2-test). Then, the factors, which had a significant between-group difference, were selected to perform Pearson correlation analysis. The significance level alpha is 0.05 (two-tailed).
Multiple mediation analysis was conducted by the PROCESS macro in SPSS, developed by Hayes (Andrew F. Hayes, 2013). The multiple mediate models could estimate a specific indirect effect is to describe how the independent variable leads the dependent variable through the intermediate factors (Andrew F Hayes, 2017). A bootstrap method was adopted to construct a 95% confidence interval for significance testing of mediating effects.