A total of 541 obese patients were consecutively enrolled in our study. The included patients were no less than 18 years, with BMI (body mass index) ≥ 28.0 kg/m². All of the subjects received physical examinations in Beijing Chao-Yang Hospital, Capital Medical University between January 2018 and January 2019 (2). Meanwhile, 501 healthy individuals with normal weight (18.5 kg/m² ≤ BMI < 24.0 kg/m²) were recruited as controls, and they were matched with obese cases in age and gender (2). The subjects had no evidences for alcohol abuse, cardiovascular disease, thyroid dysfunction, anemia, hematological disease, chronic hepatitis/cirrhosis, biliary obstruction disease, acute or chronic infections, renal insufficiency, systemic inflammation, or cancer. Individuals in pregnancy or taking medications known to influence bilirubin, liver function, insulin, glucose, lipid or blood pressure were excluded, such as glycyrrhizic acid, ursodeoxycholic acid, potassium magnesium aspartate, “Yinzhihuang” granules, Metformin, Reserpine, Guanethidine, etc. In addition, persons meeting any one of the following conditions were also removed: serum bilirubin levels ≥ 2 times of upper limit of normal (ULN) (ULNs: TBIL: 21.0 μmol/L and DBIL : 6.8 μmol/L) or serum ALT and/or AST and/or GGT levels ≥ 3ULN (ULNs: alanine aminotransferase (ALT) : 50 U/L, aspartate aminotransferase (AST) : 40 U/L, and gamma-glutamyl transferase (GGT) : 60 U/L).

When obese individuals (BMI ≥28 kg/m²) did not conform to the following criteria, they were categorized as MHO (13): (1) elevated FBG (≥5.6 mmol/L), (2) elevated TG (≥1.7 mmol/L), (3) reduced HDL-C (<1.0 mmol/L for men and <1.3 mmol/L for women), and (4) elevated SBP (≥130 mmHg) or/and DBP (≥85 mmHg). Obese patients who had one or more of these four metabolic risk components were categorized as metabolically unhealthy obesity (MUHO) (13).

This research was performed according to the Declaration of Helsinki ethical principles. Ethics approval was given by the Institutional Review Board (IRB) of Beijing Chao-Yang Hospital, Capital Medical University. Written informed consents were obtained from all subjects.

Both health status and medical history were collected through a standard questionnaire during a face-to-face interview, including alcohol consumption, medication status, physical activity, and history of diseases. Height was measured, with an accuracy of 0.1 cm, and weight was accurate to 0.1 kg, by professional medical staff, while participants were wearing light clothing without shoes. Body mass index (BMI) was calculated as weight/body height (kg/m²). We employed a sphygmomanometer to measure sitting blood pressure for non-dominant arm after at least ten-minute rest.

After overnight fasting, blood samples were collected from median cubital vein. All biochemical indicators were measured using an automatic biochemical analyzer (Hitachi 747, Roche Diagnostics, Germany), except for fasting plasma insulin (FINS), glycated hemoglobin A1c (HbA1c) and high-sensitivity C-reactive protein (hsCRP). Alanine aminotransferase (ALT), aspartate aminotransferase (AST) and gamma-glutamyl transpeptidase (GGT) were measured via velocity method. Total bilirubin (TBil) and direct bilirubin (DBil) were measured through vanadate oxidation method. Indirect bilirubin (IBil) was calculated based on TBil minus DBil. Total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) were assessed adopting colorimetric enzymatic methods. Plasma TC and TG were tested employing enzymatic cholesterol oxidase reaction and glycerol lipase oxidase reaction, respectively. HDL-C and LDL-C were assessed via direct measurement. Fasting plasma glucose (FBG) were estimated utilizing glucose oxidase assay. FINS concentrations were tested using chemiluminescence assay (Dimension Vista, Siemens Healthcare Diagnostics, Germany). HbA1c was measured applying high-performance liquid chromatography (HPLC) on an automatic biochemical analyzer (HLC-723G7, Tosoh Corporation, Tokyo, Japan). hsCRP levels were determined with immunonephelometric analysis. To estimate insulin resistance, homeostasis model assessment of insulin resistance (HOMA-IR) was calculated: HOMA-IR = [FINS (μIU/mL) * FBG (mmol/L)/22.5] (14).

All statistical analyses were performed using SPSS 21.0 (SPSS, Chicago, IL, USA). Continuous variables in normal distribution were recorded as mean values ± standard deviation (SD). Because values for ALT, GGT, TG, FINS, hsCRP and HOMA-IR did not conform to normal distribution, they were represented by the median, 25% quartile and 75% quartile. Chi-squared test was adopted to analyze categorical variables. ANOVA or Kruskal-Wallis H test was used to compare differences among the three groups. Post hoc analyses were performed. The relationships between two parameters were analyzed using Pearson and Spearman correlations. The relationship between bilirubin and other variables was analyzed using Multivariate regression analysis. P < 0.05 (two-tailed) revealed statistical significance of results.

Baseline characteristics of the included subjects were summarized in Table 1. Analysis results showed that age and gender were similar among control, MHO and MUHO groups. Among these three groups, no significant differences were observed in serum IBil levels. Significant differences were identified in BMI, SBP, DBP, ALT, AST, GGT, TBil, DBil, TC, TG, HDL-C, LDL-C, FBG, FINS, HbA1c, hsCRP and HOMA-IR levels among control, MHO and MUHO groups (all P < 0.01).

Further post hoc analysis showed that both of MHO and MUHO groups had significantly increased BMI, FINS, hsCRP and HOMA-IR compared with control group (Table 1). Moreover, patients in MUHO group had significantly increased SBP, DBP, ALT, AST, GGT, TC, TG, LDL-C, FBG and HbA1c levels, and decreased Tbil, DBil and HDL-C levels when compared to healthy controls (Table 1). MHO and MUHO groups showed similar tendencies in BMI, ALT, AST, GGT, IBil, TC, LDL-C and FINS levels (Table 1). Interestingly, MHO group had relatively lower SBP, DBP, TG, FBG, HbA1c, hsCRP and HOMA-IR levels, and higher TBil, DBil and HDL-C levels than MUHO group (Table 1).

We compared TBil, DBil and IBil levels among control, MHO and MUHO groups. As displayed in Supplementary Figure 1, MUHO group showed significantly lower level of TBil than control (P<0.01) and MHO (P<0.05) groups. MHO and control groups showed similar TBil values (P>0.05). As for IBil level, there were no significant differences among control, MHO and MUHO groups (P>0.05 for all) (Supplementary Figure 2). In addition, the level of DBil was decreased in MUHO group, compared with control and MHO groups (P<0.01 for both). Meanwhile, control and MHO groups showed insignificant difference in DBil level (P>0.05) (Supplementary Figure 3).

TBil level showed negative association with BMI, GGT, TC, TG, LDL-C, FINS, hsCRP and HOMA-IR, and positive association with HDL-C level (Table 2). Moreover, these significant correlations did not show remarkable alteration after adjusted for age, gender and BMI (P <0.01 for all).

Serum DBil level was negatively correlated with age, BMI, ALT, GGT, TC, TG, LDL-C, FBG, FINS, hsCRP and HOMA-IR, and positively associated with HDL-C levels (Table 2). After adjusted for age, gender and BMI, the associations of DBil level with TC, LDL-C, FINS, hsCRP and HOMA-IR were still significant (TC: r = - 0.493; LDL-C: r = - 0.530; FINS: r = - 0.441; hsCRP: r = - 0.335; HOMA-IR: r = - 0.380; all P <0.01).

Serum IBil level was negatively correlated with hsCRP and HOMA-IR, and positively correlated with HDL-C and LDL-C levels (Table 2). However, these associations were insignificant after adjusted for age, gender and BMI.

Multivariate regression analysis was used to evaluate the correlation between serum bilirubin level and clinical and biochemical parameters, including age, gender, BMI, TG, FBG, hsCRP and HOMA-IR. Results showed HOMA-IR was independently correlated with TBil and DBil levels (β = -0.400, P < 0.01).