In this large population-based study, we analyzed the correlation of leptin with other clinical characteristics and biochemical factors, including age, BMI, WC, blood glucose, HOMA-IR, blood lipid profiles, and UA. Furthermore, we proposed formulas for serum leptin concentrations using multivariate linear regression analysis in both men and women. We established sex- and BMI-specific reference intervals for serum leptin levels in both sexes, based on a healthy Chinese population.
Previous studies have demonstrated a relationship between leptin levels and body fat distribution in various ethnic populations. A cross-sectional study found high correlations between leptin levels and BMI (r = 0.80 in men, 0.79 in women), and concluded that leptin production was proportional to adipose tissue mass . Furthermore, another study indicated that leptin concentrations were significantly correlated with BMI (r = 0.741 in men, r = 0.814 in women) and WC (r = 0.840 in men, r = 0.718 in women) in Mexican Americans, and concluded a high association between leptin levels and overall adipose tissue depots rather than with a specific fat depot . Our study found a similar correlation between serum leptin concentrations with both BMI (r = 0.698 in men and r = 0.626 in women) and WC (r = 0.712 in men and r = 0.554 in women) for both sexes. However, some studies have reported a significant correlation between leptin with other variables of body adiposity, such as skinfold thickness , percentage body fat  and truncal body fat .
In addition to physical measures, we investigated the association between leptin and biochemical variables. Our study showed a correlation between serum leptin concentrations and HOMA-IR in a Chinese population, which is consistent with the data of a previous study, indicating that leptin levels were positively associated with HOMA-IR (P < 0.001) in South Asian, Chinese, Aboriginal, and European Canadians . Interestingly, a previous study showed a genetic correlation (r = 0.4785) and a shared genetic locus between circulating leptin concentrations and HOMA-IR . Obesity leads to increased leptin levels and insulin resistance , and an experimental study indicated that leptin regulates the development of insulin resistance through its effects on the liver . Elevated serum TG levels may lead to the expansion of adipose tissue, which may further increase serum leptin concentrations. In this study, we found that serum leptin levels were positively correlated with TG concentrations. This finding is consistent with the data of a large population-based study in Caucasian individuals, showing a significant correlation between leptin and TG:0.32 in women (P < 0.001) and 0.28 in men (P < 0.001) . In addition, animal studies have suggested that leptin may influence serum triglyceride levels indirectly [26, 27].
Based on the significant correlation of leptin with BMI, HOMA-IR, TG, and UA, we proposed the predicted formulas for serum leptin levels for each sex, and the adjusted R2 of the formula was 0.552 and 0.447 for men and women, respectively. The multiple linear regression models in our study explained the degree of variance in serum leptin levels, as high as that explained by other models, including percentage body fat (R2 = 0.523 for men; R2 = 0.551 for women) estimated by the bioimpedance method  or total abdominal fat (R2 = 0.493 for men) assessed by abdominal CT .
To the best of our knowledge, our study is the first to establish sex- and BMI-specific reference intervals for leptin for both sexes in the largest Chinese cohort. In addition, this cohort was a random sampling population instead of the hospital population. A previous Chinese study investigating the reference values of serum leptin in women indicated that the mean ± SD value of serum leptin concentrations was 10.5 ± 1.99 ng/ml , which was significantly lower than the serum levels of leptin (20.92 ± 12.96 ng/mL) in women in our study . The lower BMI levels (22.69 ± 3.26 kg/m2) observed in the above study  compared to that of our study (24.6 ± 3.5 kg/m2) may account for the variance of leptin concentrations. In addition, we established sex- and BMI-specific reference intervals for serum leptin concentrations in both sexes. A study of the data from the third National Health and Nutrition Examination Survey (NHANES III) showed the serum leptin concentrations by ethnicity and BMI in both genders , which had similar levels of leptin in men of all BMI groups (6.0 ± 0.16 ng/mL) as observed in our study. However, the serum leptin levels of all BMIs, < 20, 20 to < 25, 25 to < 27.5, 27.5 to < 30, and > 30 groups in women in our study seemed to be higher than that in the NHANES III study . Another study in multiple ethnicities also found that compared with European women, South Asian women had strikingly high leptin concentrations, which could not be explained by fat distribution and hyperinsulinism . However, there is evidence that hypertrophy of subcutaneous adipocytes may contribute to high leptin concentrations in South Asians .
After determining the reference interval of leptin levels, our further analysis showed that, in the normal BMI participants, individuals with leptin levels higher than the 97.5th percentile may have a higher risk of prevalent metabolic syndrome and abdominal obesity. This finding contributes to the clinical application of serum leptin measurement, but future studies with larger sample sizes are needed to support this finding.
The established reference intervals in our study provide reliable evidence for the clinical analysis and understanding of leptin, as well as a theoretical basis for the clinical application of leptin. However, our study had some limitations. First, the reference intervals for serum leptin concentrations can only reflect healthy individuals in the Han population. Second, we did not assess other specific fat distributions, such as skinfold thickness and visceral fat. Third, after excluding the participants who meet the exclusion criteria, there are differences in the number of participants with different age and sex. We excluded the participants with diabetes, so the higher prevalence of diabetes in men than in women may contribute to the differences. Of note, the sample size in our study is larger than previous study . In addition, the median (25th, 75th) of serum leptin absolute levels with different ages are similar (Table 3).