In the currents study, we examined the temporal sequences between obesity measures and insulin, HOMA-IR or HOMA-%β using the cross-lagged panel analysis model. The major finding of this research was that we found a bidirectional relationship between obesity and hyperinsulinemia, and the causal effect of insulin, HOMA-IR or HOMA-%β on abdominal obesity measures, especially WHR, were stronger than the inverse effect of abdominal obesity measures, while the effect of BMI on insulin, HOMA-IR or HOMA-%β was stronger than the inverse effect of insulin, HOMA-IR or HOMA-%β suggesting abdominal obesity measures are more sensitive to hyperinsulinemia measures than BMI. These findings proved a new sight on examining the unappreciated role of hyperinsulinemia in obesity and highlight the impact of general and abdominal obesity measures on the temporal relationship between obesity and hyperinsulinemia.
Although the close inter-correlation between obesity and hyperinsulinemia has been well documented, the causal relationship between them remains a point of research2, 8, 22, 27. The fact noted in this study that excessive obesity was a predictor of increased insulin and the inverse effect of insulin on obesity was also exist, indicated there was a bidirectional causal relationship between obesity and hyperinsulinemia. In previous studies, many researchers have reported the causal role of obesity on insulin, by cross-section, longitudinal or Mendelian randomization studies2, 27, 28. It is widely considered that excess BMI or obesity can cause hyperinsulinemia through multiple mechanisms such as dysregulation of lipid and glucose metabolism, inflammation, hormone imbalance, genetic variants, and so on5, 7-9. In contrary, the inverse effect of insulin on obesity also has been reported. Many studies reported the treatment of insulin led to subsequent obesity in diabetic patients10, 11. And the significant weight loss was observed in patients treated by the inhibitors of insulin secretion, diazoxide or octreotide, after the marked decrease in insulin levels12, 29, 30. Further, in recent series studies, the fact that mice with the disruption of insulin secretion or fat-specific insulin receptor genes were prevented from high-fat diet induced adiposity, providing directly genetic evidence on the inverse effect of hyperinsulinemia on weight homeostasis23, 31-33. The notable “ticking clock” hypothesis also provides a plausible explanation for the insulin-induced weight gain34. Moreover, in large longitudinal studies, several researchers had reported the prediction of hyperinsulinemia on subsequent obesity in humans3, 17, 32. Recently, Astley et al study also provided a piece of convincing evidence on the causal role of insulin on obesity by using Mendelian randomization analysis4. Taken together, evidence from our study and others indicated that the regulation of insulin on obesity was also existent, there was a bidirectional causal relationship between obesity and hyperinsulinemia.
Notably, we also observed a bidirectional relationship between multiple obesity measures and insulin resistance (HOMA-IR and HOMA-%β). One potential explanation was that participants enrolled by our study were not diabetics and had normal glucose tolerance and beta cell function, the excessive obesity stimulated the compensatory of insulin hypersecretion and resulted in the development of hyperinsulinemia and insulin resistance, then the increased insulin level, in turn, promoted later life obesity by impacting energy homeostasis and fat deposition31, 35. The Compensatory of increased beta cell function caused by excess obesity in this study and the observation of weight gain in insulin resistance patients without diabetes mellitus in previous researches do support our hypotheses3, 36.
As the surrogate for general obesity, BMI is widely used to study the relationship between obesity and hyperinsulinemia. However, several evidence suggest that WHR is also a valuable choice for investigating the link between obesity and insulin, by providing additional information about abdominal adiposity disruption21, 22. In this study, we examined the impacts of obesity measures on the temporal relationship between obesity and hyperinsulinemia, and found that the effects of insulin on abdominal obesity measures, especially WHR, were stronger than the inverse effects of abdominal obesity measures, while the effect of BMI on insulin was stronger than the inverse effect of insulin. HOMA-IR and HOMA-%β shown similar patterns. Our observations indicated that abdominal obesity measures were more sensitive to insulin levels than BMI. In biochemical studies, many researchers have reported the function of insulin on inhibiting lipolysis and promoting lipid accumulation in white adipose tissue, which is widely distributed in body subcutaneous and viscera tissue15, 32. In animal models, the Ins1+/+: Ins2-/- mice were observed to exhibit a robust increase in fat-lean rate and size of adipocytes when compared to Ins1+/-: Ins2-/- mice, whose insulin secretion were suppressed by insulin genes knockout31. Similarly, in recent Page et al study, with a significant reduction in visceral fat storage and no changed lean mass, diet-induced mice obesity was reversed by a modest inhibition of circulation insulin23. Several studies also have reported the effect of insulin hypersecretion on skinfold thickness in humans16, 37. Together, these results indicate the abdominal obesity might be more susceptible to insulin levels than general obesity. To date, no studies have examined the temporal relationship between obesity and hyperinsulinemia, involving both general and abdominal obesity measures in a longitudinal cohort of Chinese adults, and underscore the impacts of multiple obesity measures on the temporal sequences between obesity and hyperinsulinemia. Furthermore, this study also shown similar temporal patterns between obesity and hyperinsulinemia by sex groups with those findings of the total sample. The path coefficients of baseline WC or WHR to follow-up HOMA-%β were significant greater in men group than in women group, but the mechanisms were still unclear.
The major strength of this study was the use of cross-lagged panel analysis, which is a powerful method for dissecting the temporal sequences between inter-correlated variables. Another advantage is that we included all the general and abdominal obesity measures in one study and examine their impacts on the temporal sequences between obesity and hyperinsulinemia at the same time. There were some limitations of this study also need to be noted. Firstly, the definition of diabetes and drug information were only according to the health questionnaires. The potential mission of disease and treatment information might exist. Secondly, this study only recruited Chinese adults. These findings of our study may not be generalizable to other ethnic populations.