High Alanine Aminotransferase Is Associated With Decreased Hepatic Insulin Sensitivity and Predicts the Development of Type 2 Diabetes

Subjects in this study were participants in a longitudinal study of the pathogenesis of type 2 diabetes initiated in 1982 (8), and, therefore, they partially overlap with all prospective studies of type 2 diabetes in Pima Indians. All participants were Pima (or closely related Tohono O’odham) Indians from the Gila River Indian Community near Phoenix, Arizona. Subjects were invited at approximately annual intervals for repeat 75-g oral glucose tolerance tests (OGTTs) and an assessment of insulin sensitivity and insulin secretion. All subjects were between 18 and 50 years of age, nondiabetic (normal glucose tolerance [NGT] or impaired glucose tolerance [IGT]) at baseline according to OGTT (1985 World Health Organization criteria) (9), nonsmokers at the time of the study, and, except for obesity, healthy according to a physical examination and routine laboratory tests. By protocol, subjects with hepatic enzyme concentrations higher than three times the upper limit of the reference range were not studied. None of the patients had known history of hepatitis, but this was not confirmed by laboratory tests in all cases. No subject had clinical or laboratory signs of acute or chronic infection or took any medication at the time of the study. The protocol was approved by the Tribal Council of the Gila River Indian Community and by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases. All subjects provided written informed consent before participation.

Cross-sectional analyses were carried out in 451 nondiabetic (286 NGT/165 IGT) Pima Indians, 173 women and 278 men, who had been characterized for ALT, AST, and GGT concentrations; body composition; glucose tolerance; fasting plasma insulin concentrations; insulin action; hepatic glucose output (HGO) during low-dose insulin infusion; and acute insulin response (AIR) (Table 1). Plasma alkaline phosphatase, bilirubin, and albumin concentrations were available in 267 subjects (163 women and 104 men). Total white blood cell count (WBC) was available in 111 subjects (69 women and 42 men, mean 8,288 ± 1,883 cells/mm³).

Prospective analyses were performed in 370 Pima Indians with NGT at baseline, who had baseline plasma measurements of ALT, AST, and GGT, body fat, insulin action, and AIR and had a follow-up OGTT. Among them, 63 subjects developed diabetes (progressors), whereas 307 remained nondiabetic (nonprogressors), with an average follow-up of 6.9 ± 4.9 years (range 1 month to 18 years).

Longitudinal analyses were performed in 224 nonprogressors with NGT at baseline who were nondiabetic at follow-up and had a baseline and follow-up assessment of ALT, AST, GGT, body composition, fasting plasma insulin concentrations, insulin action, HGO, and AIR.

All subjects were admitted for 8–10 days to the National Institutes of Health Clinical Research Unit in Phoenix, Arizona, and were fed a weight-maintaining diet (50, 30, and 20% of daily calories provided as carbohydrate, fat, and protein, respectively) and abstained from strenuous exercise.

Plasma hepatic enzyme, albumin, and bilirubin concentrations were evaluated on admission after an overnight fast and measured by colorimetric method (DADE Behring-Dimension Clinical Chemistry System) in the hospital laboratory. WBC was measured on admission in the hospital laboratory by an automated cell counter.

Body composition was estimated by underwater weighing, with simultaneous determination of residual lung volume by helium dilution (10) or by total-body dual energy X-ray absorptiometry (DPX-L; Lunar Radiation, Madison, WI) with calculations of percent body fat, fat mass, and fat-free mass as previously described (11). Waist and thigh circumferences were measured at the umbilicus in the supine position and the gluteal fold in the standing position, and waist-to-thigh ratio (WTR) was calculated as an index of body fat distribution.

At least 3 days after admission and after a 12-h overnight fast, subjects underwent a 2-h, 75-g OGTT (8). Plasma glucose concentrations were determined by the glucose oxidase method (Beckman Instruments, Fullerton, CA) and plasma insulin concentrations by an automated immunoassay (Access; Beckman Instruments).

Insulin action was assessed at physiologic and supra-physiologic insulin concentrations during a two-step hyperinsulinemic-euglycemic glucose clamp as previously described (12). In brief, after an overnight fast, a primed continuous intravenous insulin infusion was administered for 100 min at a constant rate of 290 pmol (40 mU) per meter squared body surface area per minute (low dose, M-low), followed by a second 100-min infusion at a rate of 2,900 pmol (400 mU) · m⁻² · min⁻¹ (high dose, M-high). These infusions achieved steady-state plasma insulin concentrations of 954 ± 318 and 17,574 ± 6,486 pmol/l (means ± SD), respectively. The rate of total insulin-stimulated glucose disposal was calculated for the last 40 min of the low- (M-low) and high-dose insulin infusions (M-high). The rate of endogenous glucose output was calculated before insulin infusion [BGSO, calculated from Steele’s equation using a primed (30 μCi) continuous (0.3 μCi per min) 3-³H-glucose infusion] and during the last 40 min of low-dose insulin infusion (HGO). HGO was used as an estimate of hepatic insulin sensitivity. M-low was also corrected for the rate of HGO (12). M values were adjusted for the steady-state plasma glucose and insulin concentrations and were normalized to the estimated metabolic body size (EMBS = fat-free mass + 17.7 kg) as previously described (12).

AIR was measured in response to a 25-g intravenous glucose tolerance test and calculated as the average incremental plasma insulin concentration from 3–5 min after the glucose bolus (12). Because even mildly elevated glucose concentrations can secondarily affect insulin secretion, only data from subjects with NGT were included in the analyses performed with AIR.

Statistical analyses were performed using the software of the SAS Institute (Cary, NC). Results are given as means ± SD (unless indicated otherwise). The values for hepatic enzymes, fasting insulin, M-low, and AIR were logarithmically transformed before analysis to approximate normal distributions.

In the cross-sectional analyses, the relationship between hepatic enzymes and anthropometric and metabolic variables was examined by Spearman correlation coefficients. Multiple linear regression models and partial correlations were used to examine the relationships after adjusting for covariates. Differences between subjects with NGT and IGT were assessed by unpaired t test.

In the prospective analyses, we assessed the metabolic predictors of diabetes. Risk factors for progression from NGT to diabetes were estimated by proportional hazards analyses. Effects of hepatic enzymes, M-low, HGO, AIR, and insulin were expressed as relative hazards with 95% CIs. For the purpose of presentation, the relative hazard estimates were scaled for comparison at the 10th and 90th percentiles, i.e., the risk of developing diabetes of a hypothetical subject at the 90th percentile compared with the hazard of a hypothetical subject at the 10th percentile (or at the 10th and 90th percentiles in the case of a negatively related variable). Differences in anthropometric and metabolic variables between progressors and nonprogressors were assessed by unpaired t tests.

In the longitudinal analyses, the predictive effect of baseline hepatic enzyme concentrations on change (follow-up adjusted for baseline) in body fat, fasting plasma insulin concentrations, M-low, AIR, and HGO were evaluated using multiple linear regression models. Models were adjusted for sex, follow-up age, and the time of follow-up. Change in fasting insulin concentrations, M-low, and HGO were additionally adjusted for change in body fat, and change in AIR was adjusted for change in body fat and M-low. Differences between anthropometric and metabolic variables at baseline and follow-up were assessed by paired t tests.

The anthropometric and metabolic characteristics of the study populations for the cross-sectional, prospective, and longitudinal analyses are summarized in Table 1.

ALT, AST, and GGT concentrations did not differ between individuals with NGT and IGT (all P > 0.05). ALT was positively correlated with AST (r = 0.64, P = 0.0001) and GGT (r = 0.58, P = 0.0001), and AST was correlated with GGT (r = 0.56, P = 0.0001).

In general, after adjusting for age and sex, hepatic enzyme concentrations were positively correlated with measures of adiposity, fasting plasma insulin, and HGO and negatively correlated with the rate of insulin-stimulated glucose disposal (Table 2 and Fig. 1). Hepatic enzymes were not related to measures of central adiposity (waist and WTR, all P > 0.1) independent of body fat. After adjustment for age, sex, and adiposity, ALT, AST, and GGT were associated with M-low (all P < 0.0001). When M-low, fasting insulin concentrations, and HGO were included in the same model, all three were independent determinants of ALT concentrations (all P < 0.01). This was in contrast to AST or GGT concentrations. Only ALT was associated with M-high (P = 0.0001) after adjustment for age, sex, and body fat. ALT, AST, and GGT were positively related to WBC, before and after adjustment for body fat and insulin action, in a partial correlation (r = 0.25, 0.26, and 0.20, respectively, all P < 0.01).

Hepatic enzymes at baseline did not differ between progressors and nonprogressors (Table 1). In a proportional hazards analyses, with adjustment for age and sex, a high ALT predicted the progression from NGT to diabetes [relative hazard 90th vs. 10th percentiles (95% CI): 2.3 (1.4–3.6), P = 0.0004]. Ninetieth and tenth percentiles of ALT were 70 and 12 mU/l, respectively. The predictive effect of a high ALT persisted after additional adjustment for percent body fat, AIR, and M-low [1.9 (1.1–3.3), P = 0.02] and/or HGO [1.9 (1.1–3.3), P = 0.03]. In the model that included age, sex, percent body fat, M-low, AIR, and ALT in addition to ALT, M-low [5.2 (1.9–14.4), P = 0.003] and AIR [4.4 (1.9–9.8), P = 0.0001] were both predictive of type 2 diabetes. When subjects with IGT at baseline were included in the analysis together with subjects with NGT, ALT (adjusted for age, sex, percent body fat, M-low, and AIR) was predictive of future diabetes [2.5 (1.7–3.7), P = 0.0001]. HGO was higher in progressors compared with nonprogressors (Table 1) but was not predictive of diabetes independently of M-low and AIR [1.6 (0.95–2.8), P = 0.07]. When divided into tertiles, subjects in the upper tertile of relative ALT (≥59 mU/l, adjusted for age, sex, body fat, M-low, and AIR) had a higher cumulative incidence at all time points of follow-up compared with subjects in the lower tertile (≤21 mU/l) (Fig. 2). AST, GGT, alkaline phosphatase, bilirubin, and albumin concentrations as well as AST/ALT ratio did not predict development of type 2 diabetes (data not shown).

A high ALT at baseline was associated with an increase in HGO after adjustment for time of follow-up, follow-up age, sex, and change in body fat (P = 0.001) (Fig. 3). ALT was not associated with decline in M-low (P = 0.3). Plasma AST and GGT at baseline were not associated with change in M-low or HGO (data not shown). Plasma ALT, AST, and GGT concentrations at baseline were not associated with the subsequent change in AIR, fasting plasma insulin concentrations, or percent body fat (all P > 0.1).

In the present study, we found that among Pima Indians with NGT, high ALT concentrations predicted the development of type 2 diabetes. Moreover, we demonstrated that a high ALT at baseline was associated with a decline in hepatic insulin sensitivity.

The cross-sectional relationships between hepatic enzymes and adiposity confirm previous findings in other populations (4–6). In accordance with two previous studies, we found that the relationship between hepatic enzyme concentrations and insulin sensitivity was independent of the degree of adiposity (5,6).

Prospective analyses revealed that higher ALT concentrations predict type 2 diabetes in subjects with NGT. We have also shown that the predictive effect of ALT was independent of the degree of adiposity. Our results are in agreement with the finding of Ohlson et al. (3), who described a higher ALT as a risk factor for type 2 diabetes independent of obesity, body fat distribution, plasma glucose, lipid, AST, bilirubin concentrations, and family history of diabetes in nondiabetic Swedish men. Recently, a report by Perry et al. (4) described an association between elevated GGT and risk of developing type 2 diabetes independent of obesity and alcohol intake.

Why might ALT be associated with later development of type 2 diabetes? We further examined this by assessing whether ALT was associated with deterioration of risk factors for type 2 diabetes (obesity, M, and AIR) that have been established in the Pima population. Importantly, ALT was not associated with a decline in insulin sensitivity or insulin secretion. We have made the important and novel observation that ALT appears to have associations with both hepatic insulin resistance and later decline in hepatic insulin sensitivity. In contrast, AST and GGT concentrations at baseline were not related to a decrease in hepatic or whole-body insulin action or AIR. The etiologic role of the liver in later development of type 2 diabetes in humans is debated. The potential for isolated hepatic insulin resistance to lead to more generalized abnormalities of glucose tolerance is supported by animal models in which disruption of insulin signaling targeted to the liver causes diabetes (2).

In cross-sectional studies, high ALT has also been associated with fatty liver disease (5), and fatty liver disease has been associated with insulin resistance (13,14). The most obvious explanation for our findings is that raised ALT reflects fatty change in the liver and that this in turn reflects pathophysiological changes predating the development of type 2 diabetes. In animal models, chronic hyperinsulinemia is found to predispose the liver to both relative resistance to insulin action on suppression of gluconeogenesis, characterized by failure to increase insulin receptor substrate-2 and an increase in lipogenesis due to upregulation of sterol regulatory element–binding protein 1c (15). By this mechanism, it is proposed that hyperinsulinemia might directly lead to hepatic insulin resistance with associated fatty changes in the liver. Moreover, because insulin suppresses genes encoding gluconeogenic enzymes (16), and ALT is a gluconeogenic enzyme, it is also possible that ALT is an indicator of impaired insulin signaling not necessarily associated with liver injury. Methodological limitations in estimating liver glucose output using radiotracer techniques may explain why ALT, but not more direct measures of hepatic insulin sensitivity (such as HGO), predicts future type 2 diabetes in Pima Indians.

A second pathophysiological mechanism is also possible. We and others have been interested in the association of markers of chronic inflammation and later type 2 diabetes (17,18). Fatty change is a characteristic response of liver to the proinflammatory cytokine tumor necrosis factor-α (19). Fatty change and concomitant raised ALT may therefore reflect inflammation, which may impair insulin signaling both in the liver and systemically. Our observation that a high ALT is associated with a high WBC independent of obesity and insulin resistance is consistent with this hypothesis.

We do not know if alcohol consumption and/or hepatitis C may have contributed to the variability of ALT in this group of Pimas. However, alcohol consumption appears to be an unlikely confounder of our results, since studies have demonstrated higher, lower, and no differences in the prevalence of type 2 diabetes in alcohol consumers (20–22). Moreover, AST-to-ALT ratio and GGT, which are thought to be better indicators of alcoholic liver disease than ALT (23,24), were not predictive of type 2 diabetes in our study, and there is no relationship between self-reported alcohol consumption and type 2 diabetes in Pima Indians (25). Hepatitis C is a known independent predictor of type 2 diabetes (26–28), even in patients without cirrhosis (26). However, all of the recruited subjects were without a known history of hepatitis B and C and without clinical and laboratory signs of acute or chronic infection. Moreover, hepatitis C has only 2% prevalence among Pima Indian population (C. Wilson, unpublished observations).

In summary, high ALT concentrations are cross-sectionally associated with obesity, whole-body and hepatic insulin resistance and prospectively associated with a decline in hepatic insulin sensitivity and risk of type 2 diabetes. Our findings indicate that higher ALT is a risk factor for type 2 diabetes and indicate a potential role of increased hepatic gluconeogenesis and/or inflammation in the pathogenesis of type 2 diabetes.

The authors wish to thank members of the Gila River Indian Community for their participation. We also gratefully acknowledge Mike Milner, the nurses of the Clinical Research Unit, and the staff of the metabolic kitchen for their care of the patients in these studies.