The present study showed a robust association between low serum amylase and BMI, as well as latent associations with decreased PI at 0 and 60 min, HOMA-R, and HOMA-β, and increased QUICKI in asymptomatic subjects without advanced diabetes. Notably, the later associations were not detected in earlier larger studies [3, 5]. However, as in the previous studies, BMI was the most predominant factor explaining the serum amylase levels in linear and nonlinear manners. It is particularly interesting that adjustment for BMI and clinical confounders was necessary to unveil these latent associations in the multivariate logistic analysis. Indeed, this study revealed that, after adjusting for BMI, serum amylase was positively correlated with HOMA-R and FPI, which is attributable to the strong correlation between HOMA-R and FPI detected in this study. Since such high correlations are commonly observed in non-diabetic individuals [16, 17], this finding is not contradictory.
Nonlinear associations between low serum amylase and insulin resistance
The current results also showed complicated nonlinear associations between low serum amylase and high HOMA-R and high leptin (Figures 1 and 2), as low serum amylase was associated with severe insulin resistance but not moderate insulin resistance. A plausible explanation is that high insulin secretion and/or severe insulin resistance may cause insulin inactivity by downregulating insulin receptor expression or inhibiting insulin signaling in certain cells, including pancreatic acinar cells [18], eventually resulting in reduced insulinotropic action on the acinar cells [19]. By contrast, light to moderate insulin resistance is compensated for by increased insulin secretion, or moderately increased insulin secretion does not accompany insulin resistance, both of which lead to increased insulinotropic action and possibly increased serum amylase. Consistently, a similar nonlinear relationship was observed between serum amylase and circulating leptin levels. Increased circulating leptin, a marker for leptin resistance, is thought to be associated with insulin resistance [6, 7] and metabolic syndrome [20].
It is noteworthy that similar nonlinear associations between serum amylase and HOMA-R and leptin were observed in obese individuals (Figures 1B and 2B), Furthermore, the significant difference in serum amylase between lean and obese subjects was particularly evident in the normal leptin group (Figure 2B). Similar trends were also observed in the normal HOMA-R group, albeit these were not significant (Figure 1B). Additionally, HOMA-R and leptin were broadly similar between lean and obese subjects in the normal leptin group and the normal HOMA-R group (Figures 1C and D, 2C and D). Then, although these trends should be confirmed in large studies because of small sample sizes, these findings suggest that other factors associated with high BMI, but not insulin resistance, may also contribute to the pathophysiology of low serum amylase in non-insulin-resistant individuals. As we proposed in a recent article [5], low serum amylase might reflect a physiological response to over-nutrition with reduced food absorption to regulate energy balance. Consistently, [Kondo et al. 21] showed that serum amylase was significantly lower in obese subjects than in lean subjects and that low serum amylase readily increased with diet therapy and weight loss, which commonly accompanies the improvement of insulin resistance, though. By contrast, the lower serum amylase levels in lean and obese subjects in the high leptin and high HOMA-R groups suggests that severe insulin resistance exceeding a threshold level may predict low serum amylase, independently of obesity. This might be related with a previous finding that insulin resistance can be associated with an increased risk of cardiovascular events independently of metabolic syndrome [22].
Associations between low serum amylase and parameters in the fasting state
The observed associations of low serum amylase with HOMA-R, HOMA-β, PI at 0 and 60 min, but not with PI at 120 min, suggest that low serum amylase may be associated with low insulin action in the fasting state and for up to 1 h after feeding, rather than the later postprandial state. This finding may help to explain the significant association between serum amylase levels and FPG, but not with HbA1c, that was observed in previous large studies [3, 5]. HbA1c is an important marker for overall circulating glucose levels, including the postprandial state. In this study, serum amylase was not associated with HbA1c, consistent with the earlier studies [3, 5]. The lack of association between serum amylase and FPG in this study may be attributable to the small sample size.
Although TNFα was proposed as a candidate factor explaining the low serum amylase levels, unexpectedly, we found no significant association between low serum amylase and high TNFα in this study. This might be attributable to the small sample size and the specific characteristics of our study population, including the small proportions of subjects with overt diabetes, hypertension, or dyslipidemia. Alternatively, central insulin or leptin might be more effective on the serum levels of proinflammatory cytokines than systemic insulin or leptin, because a recent study by [Burgos-Ramos et al. 23] showed in animal study that intracerebroventricular administration of insulin and leptin remarkably reduced the serum levels of TNFα.
Clinical relevance of low serum amylase
The clinical relevance of low serum amylase is still poorly understood. In clinical practice, a deficiency of pancreatic enzymes is known as pancreatic exocrine insufficiency (PEI), which is characterized by steatorrhea, malabsorption, and malnutrition [24, 25]. PEI is commonly observed in individuals with pancreatic diseases, such as chronic pancreatitis and cystic fibrosis. However, PEI defined as low levels of fecal elastase-1 was found in approximately 50% of patients with type 1 diabetes and 20% of patients with type 2 diabetes [26, 27]. These, and other results, indicate that low serum amylase may represent a close link between pancreatic endocrine dysfunction and exocrine dysfunction. Although it is unknown whether low serum amylase fully reflects PEI, it is possible that low serum amylase may be partially related with pathogenesis of PEI considering that insulin action is impaired in both type of diabetes. To fully elucidate this issue, further studies focusing on endocrine dysfunction and exocrine dysfunction are needed.
Limitations
Some limitations of this should be mentioned. First, the sample size was small and the statistical power may be insufficient to reliably assess associations, particularly when the subjects were divided into sub-groups. Indeed, the large odds ratios and wide 95% confidence intervals, especially after adjustment for BMI, may be attributed to the small sample size. Second, the associations between variables were assessed in a cross-sectional manner, preventing us from determining causality. Prospective longitudinal studies and clinical trials are needed to evaluate the cause–effect relationship. Third, insulin resistance was assessed by HOMA-R, which is calculated from FPI and FPG. Ideally, insulin resistance should be assessed using specific methods, such as the hyperinsulinemic–euglycemic clamp test [28] or the frequently sampled intravenous glucose tolerance test [29]. However, because of the complexity of these tests, as well as their time-consuming procedures and expense, they were not performed in this study. It is also important to acknowledge that HOMA-R and QUICKI reflect hepatic insulin resistance rather than peripheral insulin resistance [30]. Thus, the current results should be interpreted with some caution. Finally, low serum amylase was defined as serum amylase < 60 IU/l, based on the results of an earlier large epidemiological study [3]. Different thresholds for low serum amylase are likely to yield different outcomes.
In conclusion, our results suggest a robust association between low serum amylase and BMI, as well as latent associations with decreased basal insulin and basal insulin secretion, and with insulin resistance in a nonlinear manner. These findings need to be confirmed in future large studies.