Expectedly, GGT levels were strongly, positively associated with HbA1c concentrations in both men and women. The present study showed no effect modification of either bilirubin or coffee consumption on the association between GGT and HbA1c. High bilirubin was consistently associated with lower HbA1c concentrations regardless of GGT levels.
The inverse association between serum bilirubin and HbA1c itself is not a novel finding
[10, 11]. We previously reported an inverse association between bilirubin and HbA1c in the present study population
. It was also shown among dyslipidemic patients that bilirubin concentrations were lower and serum GGT was higher in patients with the metabolic syndrome as compared with those without
. Nonetheless, it is notable that HbA1c was consistently lowered in those with high bilirubin across GGT levels. A cross-sectional inverse association of bilirubin with HbA1c or type 2 diabetes mellitus does not necessarily indicate a causal relationship because hyperglycemia itself may decrease serum bilirubin due to elevated oxidative stress
. The present findings add to evidence that bilirubin is protective against deterioration of glucose metabolism. Both bilirubin and GGT levels increase in the presence of hepatobiliary diseases, but bilirubin and GGT are generally uncorrelated with each other in healthy subjects
. In the present study population, bilirubin was weakly correlated with GGT only in men.
In addition to a possible protective effect against type 2 diabetes mellitus
[16, 17], coffee has consistently been shown to be associated with lower activities of serum GGT
. Indeed, coffee consumption was strongly, inversely associated with serum GGT in the present study population as well
. Coffee did not affect the association between GGT and HbA1c in either men or women, and there was no tendency that coffee drinkers had lower HbA1c concentrations when their GGT levels were higher. The present findings are seemingly in disagreement with the previous observation that a decreased risk of type 2 diabetes mellitus associated with coffee consumption was more marked in those with higher levels of GGT
. However, the present study did not address the association between coffee and the risk of type 2 diabetes mellitus, and the findings do not necessarily exclude the possibility that the risk of type 2 diabetes is decreased with coffee consumption in high normal GGT levels.
The present study firstly addressed the association between bilirubin and coffee in combination and HbA1c concentrations. None has addressed the combined effect of bilirubin and coffee on the risk of type 2 diabetes. Because coffee and bilirubin are potent antioxidants, it can be hypothesized that coffee’s effect on glucose metabolism may be modified by bilirubin status and vice versa. The inverse association between coffee consumption and HbA1c observed among women with low bilirubin is compatible with this hypothesis, but no such association was observed in men. Careful interpretation is needed, however, because the analysis on the combination of bilirubin and coffee included four statistical tests for the trend (Table
3). There is a 15% chance that at least one of the four P values will be 0.04 or less even if there is no association between coffee and HbA1c. The combined effect of bilirubin and coffee on glucose metabolism would deserve further studies. It should be noted that coffee may exert a protective effect on glucose metabolism via mechanisms other than the antioxidant effect. In vitro studies have suggested that antioxidant coffee compounds such as chlorogenic acid and caffeic acid delay glucose absorption in the intestine, inhibit glucose output in the liver and increase peripheral glucose uptake
Although men consuming 1–3 cups of coffee per day tended to have slightly lower concentrations of HbA1c, those with the highest coffee consumption had rather elevated HbA1c concentrations, especially when their bilirubin was high. We have no prompt explanation for the latter finding. The finding may be a chance fluctuation because the number of such men was small, and alternatively may be ascribed to residual confounding factors which were not controlled for in the present analysis.
In addition to a fairly large size of the study, some advantages of the present study are noted. HbA1c and bilirubin concentrations were uniformly measured, and important confounders were taken into account. Several weaknesses should be mentioned, however. Self-reported coffee consumption suffers inaccuracy, causing measurement misclassification. Causal inference is difficult for the cross-sectional association, as discussed above. Comorbid conditions may affect serum bilirubin concentrations and coffee consumption. Therefore, we excluded participants who had life-limiting morbid conditions such as atherosclerotic disease, cancer and liver disease. A low level of participation (24%) in the survey is of another concern in interpreting the results. Finally, it should be noted that the measurement in the past year was used for GGT among one-third of the subjects. Those having high GGT levels before the survey may have been under treatment for diabetes mellitus or prediabetic condition, and thereby distorting the association between GGT and HbA1c. In fact, the association between GGT and HbA1c was slightly weaker among men, not among women, with the past measurement of GGT than among men without. The adjusted geometric means of HbA1c (%) for the above-mentioned quartile categories of GGT were 5.00, 5.05, 5.14 and 5.15 respectively in men with the recorded measurement (trend P < 10–5) and 4.98, 5.08, 5.18 and 5.28 respectively in men with GGT measurement at the survey (trend P < 10–14). However, the analysis of the subjects with GGT measurement at the survey showed almost the same results as reported above with respect to the associations with bilirubin and coffee and the effect modifications of these covariates (data not shown).