The global number of people living with diabetes was estimated to be 529 million in 2021 and projected to be around 1.31 billion by 2050, with type 2 diabetes mellitus (T2DM) accounting for 96 % [1]. The complications of T2DM could significantly diminish the quality of patients’ life and increase the susceptibility to other health problems and mortality [2]. Therefore, the development of diabetes and related complications are worldwide issues, making it critical to manage the onset of T2DM to avoid progression of complications.

Diabetes mellitus is often accompanied by dyslipidemia [3], which is characterized as high triglyceride (TG) levels, low high-density lipoprotein cholesterol (HDL-C) levels, and high low-density lipoprotein cholesterol (LDL-C), suggesting that dyslipidemia may play an additional role in the pathogenesis of diabetes [4]. Accumulative evidence showed a paradoxical association of each lipid index with the risk of T2DM [5], [6], [7], [8], [9], [10], [11], [12]. Some studies found that a very high level of cumulative exposure to lipids was associated with the risk of T2DM [5], [6], [7], [8]. However, some evidence suggested that borderline high level of cholesterol or TG was also associated with the risk of T2DM [9], [10], [11]. Even evidence showed that the levels of LDL-C was not higher in people with diabetes than matched individuals without diabetes [12]. Therefore, underlying the distinct patterns of lipid indices before the diagnosis of T2DM might carry important implications for improving disease prevention of treatment.

It was also worthy to note that most current evidence mentioned above was focused on the independent effect of separate lipid index, without consideration of the correlation and joint effects of different lipid indices simultaneously. This raised uncertainty regarding the jointly impact of TG, HDL-C, and LDL-C on the risk of T2DM. Capturing the overlap between the developments of these lipids would allow us to identify subgroups that share similar patterns of joint changes over time in the same population [13], [14]. Moreover, exploring distinct longitudinal patterns of differently combined lipid indices may help better understand the variation of lipids over time and facilitate target prevention programs on T2DM. Therefore, using data from a large community-based population cohort, this study aimed to identify the jointly longitudinal multi-trajectory of TG, HDL-C, and LDL-C over time, and examine their associations with subsequent risk of incident T2DM among Chinese adults.