High body mass index (BMI) is a reliable obesity marker which has been used in numerous studies. A large cohort study suggests that high BMI is associated with the occurrence of cardiovascular diseases (CVD) (1). Overweight and obesity affect all individuals but are especially common among people with diabetes. In addition, patients with type 2 diabetes mellitus (T2DM) have a higher risk of cardiovascular disease (CVD) than the healthy population (2, 3).

It is well known that low level of high-density lipoprotein cholesterol (HDL-C) is a cause of cardiovascular disease, but recently the effect of low-density lipoprotein cholesterol (LDL-C) on CVD gradually be noticed (4–6). A cohort study has been reported that non-HDL cholesterol levels link to long-term risk of CVD (7). Moreover, among patients with obesity, LDL-C reduction with statins decreases CVD events, while drug treatments that improve atherogenic dyslipidemia have shown modest efficacy in CVD outcomes (8). Recent mendelian randomization studies also show that LDL-C is a risk factor for CVD (9, 10). These studies suggest that LDL-C concentration and its therapeutic modifications are also closely related to CVD, compared with HDL-C and TG (11).

A National Health and Nutrition Examination Survey (NHANEs) data-based study has been reported that the association between BMI and LDL-C in a normal population, the increase of LDL-C with BMI only in lean individuals, while LDL-C gradually tended to be flat or even decreased in obese people (12). Li’s study also gives a similar conclusion and found a non-linear relationship between BMI using Chinese adult data (13). However, due to sample size limitations, no study has made recommendations for patients with diabetes. In addition, a potential sex difference in the association between LDL-C and BMI should be illustrated in detail.

Therefore, this study aimed to investigate the association between BMI and LDL-C in Chinese adults with T2DM. We hypothesized that there was a non-linear dose-response relation between BMI and LDL-C in diabetes, and the sex difference existed. Our findings may identify the high CVD risk in T2DM to improve LDL-C and provide information on the prevention and treatment of CVD in Asian populations.

In this study, we found an Inverted U-shaped association between BMI and LDL-C in male patients with T2DM, for which the LDL was increased with BMI in the relatively lean population, while LDL-C gradually tended to be flat or even decreased in the more obese population. However, the inverted U-shaped smooth curve of BMI and LDL-C was not found in female patients with T2DM. Similar results have been reported in the normal population (13, 18). However, there were no studies giving the suggestion for diabetes patients limited by sample size.

Moreover, the spline did not show an Inverted U shape in males with HbA1c > 9% compared with the group of HbA1c ≤9%, which means that the sex difference of association between LDL and BMI vanished in the population with poor HbA1c control. In fact, HbA1c is directly correlated with higher LDL, especially in people with poor blood glucose control (Supplement Figure 2). This confounding effect resulted in the disappearance of gender differences in the present study. In the future, more mechanistic studies are needed to elucidate this phenomenon.

To our knowledge, this was the first study aimed to investigate the non-linear dose-response relationship between BMI and LDL-C in diabetes, which may help to identify the high CVD risk of diabetes patients. Assessing the association between BMI and CVD risk factors can help to prevent the onset or progression of the disease and explaining the non-linear relationship between the two can help quantify the risk levels of people with different BMI levels (13, 19, 20).

The association between BMI and lipid metabolism was estimated by TG and HDL-C in the previous studies. However, there was insufficient evidence of the effect on LDL-C (21). The phenomenon of the sharp truncation between BMI and LDL-C was previously overlooked. The mechanisms of this situation may be explained by the changes in the efficiency of lipid deposition in the adipose tissue for obesity (22). In the lean population, normal fat production implies that lipoprotein lipase rapidly offloads fatty acids from triglyceride-rich fats, converting them to cholesterol-rich LDL-C (23). Therefore, weight gain is associated with an increase in these lipid components. However, when lipid deposition efficiency in adipose tissue is reduced due to reaching functional limits, triglycerides may be stagnated in VLDL (13, 24). This was consistent with the results in the present study that LDL-C remained stable or decreased with increasing BMI in the obese population.

There are some differences in the results between the normal population and diabetes patients. Compared with the second slope (>=Breakpoint) of the model between BMI and LDL-C in normal males, the slope of the model in diabetes of males was smoother (Table 3). Moreover, the value of the breakpoint of the model between BMI and LDL-C in diabetes of males was larger than the value in normal males. The reason for this may be that the average BMI of the participants in this study was higher than the normal population.

In addition, the results were similar for the age and course subgroups. The similar trend of the curve could be explained by the fact that patients with a shorter course of diabetes are younger. The Inverted U shape was more pronounced in the subgroup with a shorter duration of diabetes and younger age. However, the Inverted U shape was not found in these people, who were elder and had a long course of diabetes. Li’s studies also found the age difference in the association between BMI and LDL-C, which was consistent with the results of this study (13).

The sex difference in the association between BMI and LDL may be explained by the following reason. First, sex-specific hormones could lead to sex differences in lipid metabolism, the previous studies suggested that the level of LDL-C was associated with hormones (25, 26). The level of LDL-C is correlated with menopausal status, that is, postmenopausal females have higher concentrations of LDL-C (27). In addition, a previous study suggested that hormone replacement therapy could modify the activity of LDL receptors (28). Thus, differences in hormonal status between males and females may lead to different associations between BMI and LDL-C. Second, there are sex differences in the concentrations of some cholesterol-regulating substances, which may affect lipid metabolism. For example, there was a higher PCSK9 concentration in females than in males (29). Third, sex-specific BMI and LDL-C associations may be influenced by differences in the distribution of adipose tissue between males and females. Females store more lipids and have higher percent body fat, less visceral white adipose tissue, and more subcutaneous adipose tissue than males (30). Moreover, females have a higher rate of TG synthesis compared to males. More mechanistic studies should be done to support this point.

Although a previous study reported the linear association linear relationship between BMI and LDL-C in a normal population (30), we made some substantial contributions to the literature. First, the primary results were analyzed with a high degree of accuracy using data from a national project to manage metabolic patients and according to the same standard. Second, we addressed sex differences in the non-linear association between BMI and LDL-C using generalized additive models in diabetes patients, which could explain a detailed quantification of ascending and descending slopes through dose-response analyses. Finally, in the context of the current epidemic of diabetes, the focus of this study on diabetes has important implications for the prevention and treatment of diabetes.

This study also has several limitations. First, we can’t infer causality and must highlight reverse causality as a possibility due to the cross-sectional study design. More cohort studies should be designed to clarify the association between BMI and LDL-C in type 2 diabetes. In future research, we will analyze the association between BMI and the index of lipid metabolism in a cohort using causal inference methods. Second, we could not get the indirect obesity index (e.g., body fat percentage); it was not analyzed in the present study. Third, although the result of the association between BMI and LDL-C was adjusted by a series of sensitivity analyses, residual confounders still existed, such as the usage of hormones and other drugs. Fourth, the GAM method aims to describe the nonlinear trend between two variables. It illustrates the trend in graphs, so we can use a smaller sample size for modeling. Although the GAM method does not suggest a minimum sample for modeling, a larger sample size may make the result more powerful (31). It’s difficult to get more data on diabetes from a single center. In the future study, we will include multi-center data and expand the sample to verify the conclusions of this study. Last, we could not include the patients with type 1 diabetes mellitus who have younger ages and morbid state. More studies are needed to explore the relationship between LDL-C and BMI in type 1 diabetes.

There was an Inverted U shape association between BMI and LDL-C in male patients with T2DM, for which the LDL-C was increased with BMI in the lean population, while LDL-C gradually tended to be flat or even decreased in the obese population. However, the Inverted U-shaped smooth curve of BMI and LDL-C was not found in female patients with T2DM.

The data analyzed in this study is subject to the following licenses/restrictions: The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. Requests to access these datasets should be directed to Zhao Dong, zhaodong@ccmu.edu.cn.

The studies involving human participants were reviewed and approved by Medical Ethics Committee of Beijing Luhe Hospital, Capital Medical University. The patients/participants provided their written informed consent to participate in this study.

LK: Conceptualization, methodology, analysis, writing- original draft and editing. CB and WX: Methodology, writing-review and editing. KJ and CT: Methodology. DZ: Conceptualization, methodology, writing- review and editing, supervision. All authors contributed to the article and approved the submitted version.