This community-dwelling prospective study demonstrated that the higher HDL-C levels were independently and negatively associated with the risk of newly-onset T2DM. Notably, this adverse correlation was similar in multivariable competitive risk model when death was regarded as a competitive event. Otherwise, time-dependent HDL-C levels were significantly associated with the risk of T2DM incidence. Furthermore, there was a decreasing and non-linear dose-response relationship between the continuous HDL-C levels and the T2DM risk. All the analyses in the current study presented a robust conclusion that HDL-C levels are independent risk factors of T2DM incidence.
Previous studies have already demonstrated the relationship between HDL-C levels and the future risk of T2DM incidence after adjusting for the potential covariates [12, 14, 25–27]. The PREVENT study, based on Danish patients without diabetes at an average age of 49 years, lasted for 8 years of follow-up, showed that high HDL-C was a protective factor for the risk of T2DM incidence with an odds ratio (OR) of 0.55 (0.47–0.64) [12]. High baseline HDL-C levels were also associated with a decreased risk of new T2DM in hypertensive patients. Moreover, 14120 participants with significant predictors of incident T2DM were enrolled in The ASCOT-BPLA trial, which reported the HR to be 0.72 (95% CI, 0.58–0.89) for each unit rise in HDL-C [25]. Consistently, a longitudinal study enrolling participants from the fifth clinic examination of the Framingham Offspring study reported a negative association between the HDL-C levels and the T2DM risk with an OR being 0.96 (95% CI, 0.95–0.98) [27]. However, two Chinese studies reported the contradictory conclusions. The multivariable-adjusted model based on a rural China cohort study, demonstrated a non-association between the T2DM risk and the HDL-C levels in the fourth quartile versus the lowest quartile with a HR being 0.92 (95% CI, 0.70–1.19) [15]. Similarly, another prospective study of Chinese people also failed to independently predict T2DM incidence (OR = 0.460; P-value = 0.189) [16]. In these two Chinese studies, enrolling participants with low and moderate HDL-C levels may contribute to inconsistent results with studies based on the American or European populations. In addition, in these two Chinese studies, the participants only underwent two visits to access the medical history and examination in a long period of 7 and 15 years, respectively. The underestimate of T2DM incidence in a long time interval may have an effect on the evaluation of true association between the HDL-C levels and the T2DM risk.
Most of previous studies, including Chinese studies, focused only on a single measurement, failing to adjust time-dependent variables, which may lead to inconsistent conclusions since HDL-C concentrations usually fluctuate over time. In current study, while we implied HDL-C as a time-dependent variable, other time-varying confounders that may affect the risk of our endpoint were also used to estimate the association between time-dependent HDL-C levels and the T2DM incidence accurately. In addition, we disposed death as a competitive event to reduce the bias caused by right-censored processing of non-terminal event. Moreover, the robustness of our observations that both the baseline and time-dependent HDL-C levels were significantly associated with the risk of T2DM was verified. To the best of our knowledge, it is the only prospective study in Chinese population that evaluate the impact of HDL-C levels on the T2DM risk by the multiple analysis using the baseline and time-dependent values of variables.
Several possible mechanisms based on clinical and experimental researches have been reported to explain the association between higher HDL-C and decreased risk of T2DM incidence, including increased insulin secretion, glucose uptake by peripheral muscles, and an anti-inflammatory response [28–34]. In vitro studies have shown that HDL can counteract the damage to the insulin secretion process caused by oxidized low-density lipoproteins while a random-control trial in vivo suggested that infusion of recombinant HDL promotes the activation of AMP-activated protein kinase pathway in skeletal muscles among person with diabetes [33, 35]. In double-blinded placebo experiments, the participants who received an intravenous dose of reconstituted HDL-C (rHDL-C) showed an excellent improvement in the plasma glucose. Furthermore, the apolipoprotein (apo) A-I, an important component of HDL-C, have been shown to activate the AMP-mediated protein kinase pathway (AMPK) and hence, the glucose uptake by peripheral muscles [28, 33]. Moreover, by injecting human apoA-I into an insulin-resistant mice, a sharp increase in the insulin secretion was observed and as well as the enhanced ability to eliminate the glucose [29]. Another Study utilizing the humans and rats demonstrated that the HDL-particles could neutralize the damage caused by the oxidized LDL cholesterol to β-cell function via the JNK pathway [30]. In addition, another study on mice concluded that low HDL-C levels reduced the glucose uptake in peripheral skeletal muscles by disrupting the respiratory function of mitochondria in skeletal muscle cells [31].
There are several major strengths in this study. Firstly, our study was based on a well-designed cohort with standardized procedures and strict quality control, which may represent a sample of middle-aged and elderly people in China. The data in our study are robust and the conclusions are compelling. Secondly, we used repeated measurements of HDL-C levels as a time-variant variable to reduce the regression dilution bias and provide more accurate assessment of association between HDL-C levels and the T2DM risk. Lastly, competitive risk analysis considering death as a competing event were employed to make sure the robust conclusions of association between HDL-C levels and the T2DM risk.
We acknowledge some limitations in this study. First, lacking the 2-h post-load glucose test and fasting HbA1c might induce misclassification of T2DM cases. Therefore, some participants who were supposed to develop diabetes were not diagnosed. However, we identified diabetes according to multiple measures of fasting plasma glucose and self-report physician diagnosis in a relative short time interval, and the impact of underdiagnoses of diabetes on the observed association between HDL-C levels and the T2DM risk could be small. Second, information on dietary habits were not collected in the current study since they might have influence on the levels of HDL-C, which could confound the association between HDL-C and T2DM risk. Third, focusing on elderly population in Beijing, our study cannot fully ensure the stability of extrapolating the results to general population given the ethnic difference. Therefore, further studies based on the epidemiological and experimental methods are required to explore the pathogenesis and the mechanisms of association between the HDL-C levels and the risk of newly-onset diabetes.