Abstract
Aims/Introduction
To prospectively investigate whether simultaneous elevation of gamma‐glutamyl transferase (GGT) and alanine aminotransferase (ALT) is associated with the increase of type 2 diabetes mellitus incidence independent of alcohol drinking, body mass index and triglycerides.
Methods
A total of 2,775 Japanese male workers who had no history of type 2 diabetes mellitus were followed. High GGT and ALT were defined as the top tertiles (GGT cutpoint: 49 IU/L, ALT cutpoint: 28 IU/L). Three groups were created using these dichotomized GGT and ALT cutpoints: both low, either high or both high. Multivariable Cox proportional hazards models were carried out adjusted for potential confounding factors.
Results
A total of 276 type 2 diabetes mellitus cases were identified during 12 years (27,040 person‐years) of follow up. Participants with simultaneously elevated GGT and ALT had a significantly higher incidence of type 2 diabetes mellitus, even after adjustment for fasting insulin and fasting blood glucose compared with the group without GGT or ALT elevation. Similar associations were observed in non‐ or light‐to‐moderate alcohol drinkers, as well as in participants with normal weight. However, the association was weaker in participants with triglycerides <150 mg/dL. We then evaluated whether the addition of GGT and ALT would improve the prediction of type 2 diabetes mellitus incidence, and found that their inclusion significantly increased the C‐statistic, net reclassification improvement and integrated discrimination improvement.
Conclusions
Simultaneous elevation of GGT and ALT was significantly associated with type 2 diabetes mellitus incidence, independent of potential confounding factors, including alcohol drinking and obesity, although the association might require concomitant elevation of triglycerides. Inclusion of GGT and ALT improved type 2 diabetes mellitus risk prediction.
Keywords: Alanine aminotransferase, Gamma‐glutamyl transferase, Type 2 diabetes mellitus
Introduction
The epidemic of diabetes mellitus is a serious global public health issue1. In Japan, the prevalence of diabetes mellitus is expected to increase from 7.9 to 9.8% during the next two decades2. Understanding the pathophysiological pathways leading to type 2 diabetes mellitus would be important for planning effective and efficient prevention programs.
The liver plays an important role in the regulation of blood glucose levels, especially in the fasting state3, and liver injury indicated by the elevation of blood alanine aminotransferase (ALT) or gamma‐glutamyl transferase (GGT) levels has been reported to increase the incidence of type 2 diabetes mellitus4, 5, 6, 7. These enzymes would be markers of underlying liver pathology, such as liver fat accumulation or excessive alcohol intake8, 9, although a recent Mendelian randomization study suggested that ALT might be a causal factor10. Thus, we decided to explore the associations of ALT and GGT with the development of type 2 diabetes mellitus in a cohort of middle‐aged Japanese men, not only by statistically controlling for potential confounding factors, but also by carrying out several stratified analyses by important confounding or mediating variables, including alcohol intake, obesity and hypertriglyceridemia11, 12, 13, 14.
Furthermore, it would be clinically relevant to examine the simultaneous elevation of ALT and GGT levels, and its relationship with the development of type 2 diabetes mellitus incidence. Although previous cross‐sectional studies have shown that the combination of higher GGT and ALT had a positive association with type 2 diabetes mellitus15, 16, prospective studies on that issue are limited. As a statistically significant association does not necessarily indicate usefulness for prediction, we also evaluated the incremental predictive value of the liver enzymes.
In the present study, the association of elevated GGT and/or ALT with the development of type 2 diabetes mellitus was examined prospectively with stratification by alcohol drinking, obesity or elevated triglyceride (TG) levels. We evaluated the predictive value of adding GGT and ALT for type 2 diabetes mellitus incidence at 10 years, which has not been previously done.
Methods
Study population
The Aichi Workers’ Cohort Study established in 1997 is an ongoing prospective study of diabetes mellitus and cardiovascular disease in Aichi Prefecture, located in central Japan.
In 2002, 6648 Japanese civil servants aged 35–66 years participated in the wave 2 baseline survey by responding to self‐administered questionnaires and providing their mandatory annual health checkup data17. The present study was first restricted to 5,177 male participants because of the small number of female type 2 diabetes mellitus cases observed during the follow‐up period. The following participants were then excluded sequentially: (i) 842 who did not give consent to our use of medical history or health checkup data; (ii) 448 prevalent type 2 diabetes mellitus cases at baseline, defined as self‐reported medication use or baseline fasting blood glucose (FBG) level ≥126 mg/dL or glycated hemoglobin ≥6.5% (with the USA National Glycohemoglobin Standardization Program method); (iii) 1,108 with missing values for baseline fasting insulin concentration, FBG or smoking status; and (iv) four individuals aged ≥65 years at baseline. Thus, 2,775 participants were left for the analysis.
The study protocol was approved by the Bioethics Review Committee of Nagoya University School of Medicine, Nagoya, Japan (approval number: 504–4).
Measurement of exposures, confounders and outcomes
GGT, ALT and other laboratory measures
Blood was drawn after at least 8 h of fasting. Serum samples were collected and stored at −80°C until biochemical assay. All assays were carried out at a commercial laboratory using the standard procedures. ALT and GGT were measured by the consensus method of the Japan Society of Clinical Chemistry. TG and total cholesterol (TC) were enzymatically determined. High‐density lipoprotein cholesterol (HDL‐c) was measured by the phosphotungstate method. FBG was enzymatically determined by the hexokinase method. Insulin concentration was measured by a solid‐phase radioimmunoassay (RIABEAD II; Abbott Japan Co. Ltd., Tokyo, Japan).
Covariates
Baseline weight and height were obtained at the annual health checkup. Height was measured in the standing position to the nearest 0.1 cm, and bodyweight was measured to the nearest 0.1 kg. The body mass index (BMI) was computed as bodyweight (kg) divided by the square of height (m) and used as a continuous variable in the analyses. Information on alcohol consumption was self‐reported. Participants responded to items regarding the frequency of drinking, type of alcohol and the amount of each type of alcohol per occasion. Alcohol intake (g/day) was estimated based on the responses to these items. Non‐ or light‐to‐moderate drinking was defined as those who responded that alcohol intake was <20 g/day18. Current smoking status was dichotomized as non‐current smoker or current smoker. Physically active individuals were defined as those who engaged in a moderate or vigorous leisure‐time exercise for a total of ≥60 min for ≥4 days per month. A family history of diabetes was defined as one or more first‐degree relatives reported to have the disease.
Ascertainment of incident type 2 diabetes mellitus
The participants were followed through 31 December 2014. The person‐years were calculated from baseline to the date of censoring, ascertainment of incident type 2 diabetes mellitus or the end of the follow up, whichever came first. Participants were censored when they died or retired from the workplace, except for those who agreed to provide their health history information to the researchers after their retirement (46.7% of retirees). Incident type 2 diabetes mellitus cases were ascertained through annual mandatory health checkups at workplaces until retirement and questionnaire surveys during employment, as well as after retirement. For the former, the health checkup data were annually reviewed, and type 2 diabetes mellitus incidence was defined to occur when the FBG level first became ≥126 mg/dL or glycated hemoglobin became ≥6.5%. The glycated hemoglobin test was provided only to employees aged 40, 45, 50 and 55 years until 2007, and to those with positive urinary glucose after 2008. A self‐administered questionnaire survey was carried out approximately biennially between 2004 and 2014. Participants were asked to report their medical histories of selected conditions including type 2 diabetes mellitus. The participants who reported a type 2 diabetes mellitus history were requested to provide the detailed contact information of the physicians who took charge of their disease management. The participants’ medical records were confirmed with their physicians when participants gave their written consent.
Statistical analysis
First, participants were divided into two groups by the upper cut‐off values of ALT and GGT tertiles: low ALT 5–27 and high ALT 28–578 IU/L; low GGT 8–48 and high GGT 49–1,121 IU/L. Subsequently, participants were grouped into three groups by the combination of the above two liver markers: “both low,” “either high” and “both high.” Those with only high ALT and those with only high GGT were collapsed, as type 2 diabetes mellitus incidence in these categories was low (<10). Cox proportional hazards regression models according to the combination of ALT and GGT were carried out to estimate hazard ratios (HRs) and respective 95% confidence intervals (95% CIs) for the risk of type 2 diabetes mellitus. The first model was adjusted for potential baseline confounders, including continuous variables of age, alcohol intake (g/day), logarithmically transformed (Ln‐) BMI, Ln‐TG, Ln‐HDL‐c and Ln‐TC, as well as categorical variables of current smoking status (yes/no), regular physical activity (yes/no) and family history of diabetes mellitus (yes/no). The second model was further adjusted for Ln fasting insulin, Ln‐GGT or Ln‐ALT and a categorical variable for FBG levels (<100, 100–109, ≥110 mg/dL). Stratified analyses were carried out by the presence/absence of excessive drinking (alcohol intakes ≥20 g/day)18, overweight (BMI ≥25 kg/m2)19 and hypertriglyceridemia (TG ≥150 mg/dL), as well as by low HDL‐c (HDL‐c < 40 mg/dL), high TC (TC ≥ 240 mg/dL)20, high FBG (FBG ≥ 100 mg/dL)21 and high fasting insulin levels (fasting insulin ≥ 10 μU/mL)22.
We also evaluated the incremental predictive value of adding Ln‐GGT or/and Ln‐ALT for the 10‐year risk of type 2 diabetes mellitus incidence by the C‐statistic, continuous integrated discrimination improvement (IDI) and net reclassification improvement (NRI) with 95% CI. The C‐statistic is the area under the receiver operating curve among individuals with different event times, and the values 0.70 to <0.80, and >0.80 are judged to have acceptable and excellent discrimination, respectively23.24 NRI and IDI are alternatives to the increase in the C‐statistic for evaluating improvement in the performance of different two models by the addition of new markers24, 25. The base model included potential confounding variables listed in Table 2. Calculation of these indicators was based on 300 iterations of perturbation resampling. To evaluate the predictive value for the combination of GGT and ALT, their interaction term was also included.
Statistical analyses were carried out using IBM SPSS Statistics for Windows software, Version 24.0 (Armonk, NY, USA) and R version 3.4.3 (Vienna, Austria) for Windows (http://cran.r-project.org/). A P‐value <0.05 was considered to be significant.
Results
Baseline characteristics
The mean age of the participants in the present study was 48.1 years. Participants in the highest GGT category were older and more likely to be current smokers (Table 1). They also had significantly higher alcohol intake, BMI, and blood levels of triglycerides, fasting insulin levels and FBG. Participants in the highest ALT category were also significantly older, and had higher BMI and blood levels of triglycerides, fasting insulin levels and FBG. Alcohol intake and smoking status did not differ among categories of ALT tertile.
Table 1.
Baseline characteristics of the participants according to gamma‐glutamyl transferase and alanine aminotransferase levels, Aichi cohort study 2002, Japan
| Gamma‐glutamyl transferase (IU/L) | P‐value | Alanine aminotransferase (IU/L) | P‐value | |||||
|---|---|---|---|---|---|---|---|---|
| TI (8–27) | T2 (28–48) | T3 (49–1,121) | T1 (5–18) | T2 (19–27) | T3 (28–294) | |||
| (n = 950) | (n = 862) | (n = 963) | (n = 983) | (n = 842) | (n = 950) | |||
| Age (years) | 47.0 ± 7.3 | 48.5 ± 7.2 | 49.0 ± 6.5 | <0.01 | 48.5 ± 7.2 | 48.5 ± 7.0 | 47.5 ± 6.9 | <0.01 |
| Family history of diabetes (yes) | 124 (13.1) | 128 (14.8) | 120 (12.5) | 0.30 | 129 (13.1) | 99 (11.8) | 144 (15.2) | 0.10 |
| Regular physical activity (yes) | 207 (21.8) | 174 (20.2) | 200 (20.8) | 0.70 | 223 (22.7) | 174 (20.7) | 184 (19.4) | 0.20 |
| Current smoker (yes) | 285 (30.0) | 296 (34.3) | 403 (41.8) | <0.01 | 347 (35.3) | 285 (33.8) | 352 (37.1) | 0.36 |
| Alcohol intake (g/day) | 9.1 ± 16.2 | 14.9 ± 18.9 | 26.8 ± 26.0 | <0.01 | 16.4 ± 21.6 | 17.1 ± 22.0 | 17.7 ± 22.9 | 0.48 |
| <20 g/day | 817 (86.0) | 627 (72.7) | 474 (49.2) | <0.01 | 695 (70.7) | 589 (70.0) | 634 (66.7) | 0.14 |
| ≥20 g/day | 133 (14.0) | 235 (27.3) | 489 (50.8) | 288 (29.3) | 253 (30.0) | 316 (33.3) | ||
| Fasting blood glucose (mg/dL) | 91.6 ± 9.2 | 93.3 ± 8.8 | 95.1 ± 9.9 | <0.01 | 92.4 ± 9.4 | 93.6 ± 9.5 | 94.2 ± 9.4 | <0.01 |
| <100 mg/dL | 797 (83.9) | 681 (79.0) | 689 (71.5) | <0.01 | 810 (82.4) | 643 (76.4) | 714 (75.2) | <0.05 |
| 100–109 mg/dL | 112 (11.8) | 139 (16.1) | 189 (19.6) | 123 (12.5) | 150 (17.8) | 167 (17.6) | ||
| ≥110 mg/dL | 41 (4.3) | 42 (4.9) | 85 (8.8) | 50 (5.1) | 49 (5.8) | 69 (7.3) | ||
| Body mass index (kg/m2)† | 22.2 (22.1–22.4) | 23.3 (23.1–23.4) | 23.7 (23.5–23.9) | <0.01 | 22.1 (22.0–22.3) | 23.0 (22.9–23.2) | 24.1 (23.9–24.2) | <0.01 |
| Triglycerides (mg/dL)† | 94.9 (92.0–97.9) | 114.4 (110.5–118.5) | 145.9 (140.9–151.2) | <0.01 | 95.5 (92.5–98.6) | 116.1 (112.2–120.3) | 144.5 (139.4–149.7) | <0.01 |
| HDL‐c (mg/dL)† | 55.1 (54.2–56.0) | 53.8 (52.9–54.8) | 54.9 (54.0–55.8) | 0.12 | 57.6 (56.7–58.6) | 54.2 (53.2–55.1) | 52.0 (51.2–52.8) | <0.01 |
| Total cholesterol (mg/dL)† | 199.1 (197.2–201.1) | 207.8 (205.6–209.9) | 214.7 (212.5–216.9) | <0.01 | 199.8 (197.8–201.8) | 207.6 (205.5–209.7) | 214.5 (212.3–216.8) | <0.01 |
| Fasting insulin (μU/mL)† | 5.6 (5.4–5.9) | 6.4 (6.1–6.6) | 7.1 (6.8–7.4) | <0.01 | 5.1 (4.9–5.3) | 6.2 (6.0–6.5) | 8.1 (7.7–8.5) | <0.01 |
| GGT (IU/L)† | 19.7 (19.4–20.0) | 36.1 (35.7–36.5) | 86.4 (83.7–89.2) | <0.01 | 27.0 (26.1–27.9) | 38.1 (36.6–39.6) | 61.4 (58.7–64.3) | <0.01 |
| ALT (IU/L)† | 17.3 (16.9–17.7) | 23.3 (22.7–24.0) | 31.8 (30.7–32.8) | <0.01 | 14.0 (13.8–14.2) | 22.4 (22.2–22.6) | 41.6 (40.7–42.6) | <0.01 |
Data are reported in mean ± standard deviation or number (%), unless otherwise specified. P‐values are from analysis of variance for continuous variables, and the χ2‐test for categorical variables. †Geometric mean and the 95% confidence interval. ALT, alanine aminotransferase; GGT, gamma‐glutamyl transferase; HDL‐c, high‐density lipoprotein cholesterol; T, tertile.
Association between GGT or ALT and type 2 diabetes mellitus incidence
During the 12‐year follow‐up period of 2,775 Japanese male workers (27,040 person‐years), 276 developed type 2 diabetes mellitus. The overall crude incidence rate was 10.2 per 1,000 person‐years. It was confirmed that most of the potential confounding variables were associated with type 2 diabetes mellitus incidence, except for regular physical activity and total cholesterol (Table 2).
Table 2.
Crude hazard ratios and the 95% confidence intervals for the incidence of type 2 diabetes mellitus according to potential confounding variables, Aichi cohort study 2002–2015, Japan
| Crude HR (95% CI) | ||
|---|---|---|
| Age (years) | Per 1 year | 1.0 (1.0–1.1) |
| Family history of diabetes (yes) | 1.9 (1.4–2.5) | |
| Regular physical activity (yes) | 0.8 (0.6–1.1) | |
| Current smoking status (yes) | 1.5 (1.2–2.0) | |
| Alcohol intake (g/day) | Per 1 g/day | 1.0 (1.0–1.0) |
| Ln‐body mass index (kg/m2) | Per 1 SD | 1.4 (1.3–1.6) |
| Ln‐triglycerides (mg/dL) | Per 1 SD | 1.4 (1.3–1.6) |
| Ln‐HDL‐c (mg/dL) | Per 1 SD | 0.7 (0.7–0.8) |
| Ln‐total cholesterol (mg/dL) | Per 1 SD | 1.1 (0.9–1.2) |
| Fasting blood glucose | ||
| <100 mg/dL | REF | |
| 100–109 mg/dL | 3.2 (2.4–4.2) | |
| ≥110 mg/dL | 6.8 (5.0–9.3) | |
| Ln‐fasting insulin (μU/mL) | Per 1 SD | 1.4 (1.3–1.6) |
Total sample n = 2,775. Standard deviation (SD) for logarithmically transformed (Ln)(body mass index) = 0.1; SD for Ln(triglycerides) = 0.6; SD for Ln(high‐density lipoprotein cholesterol) = 0.3; SD for Ln(total cholesterol) = 0.2; SD for Ln(fasting insulin) = 0.7. HDL‐c, high‐density lipoprotein cholesterol; REF indicates reference.
The incidence rates became higher according to the increase in the GGT or ALT categories (Table 3). Multivariable‐adjusted HRs of type 2 diabetes mellitus for men in the highest GGT or ALT categories compared with the lowest tertile were significantly higher than unity. Further adjustment for fasting insulin, FBG and each liver enzyme attenuated the significant association of GGT with type 2 diabetes mellitus, but not for the association of ALT.
Table 3.
Crude incidence rates, multivariable‐adjusted hazard ratios and the 95% confidence intervals for the incidence of type 2 diabetes mellitus according to tertiles of gamma‐glutamyl transferase or alanine aminotransferase, Aichi cohort study 2002–2015, Japan
| Gamma‐glutamyl transferase (IU/L) | Continuous HR per 1 SD | Alanine aminotransferase (IU/L) | Continuous HR per 1 SD | |||||
|---|---|---|---|---|---|---|---|---|
| TI (8–27) | T2 (28–48) | T3 (49–1,121) | T1 (5–18) | T2 (19–27) | T3 (28–294) | |||
| (n = 950) | (n = 862) | (n = 963) | (n = 983) | (n = 842) | (n = 950) | |||
| No. incidents | 59 | 71 | 146 | 67 | 71 | 138 | ||
| Person‐years | 9,691 | 8,325 | 9,025 | 9,764 | 8,234 | 9,043 | ||
| Crude incidence rate | 6.1 | 8.5 | 16.2 | 6.9 | 8.6 | 15.3 | ||
| HR 1 | REF | 1.1 (0.8–1.5) | 2.0 (1.4–2.8) | 1.4 (1.2–1.6) | REF | 1.1 (0.8–1.6) | 1.7 (1.2–2.3) | 1.4 (1.2–1.5) |
| HR 2 | 1.0 (0.7–1.4) | 1.4 (1.0–2.1) | 1.1 (0.9–1.3) | 1.0 (0.7–1.4) | 1.3 (0.9–1.8) | 1.2 (1.1–1.4) | ||
Hazard ratio (HR) 1: adjusted for age, family history of diabetes (yes/no), regular physical activity (yes/no), current, smoking status (yes/no), alcohol consumption (g/day), logarithmically transformed (Ln) body mass index (kg/m2), Ln triglycerides, Ln high‐density lipoprotein cholesterol and Ln total cholesterol. HR 2: adjusted for same covariates as HR 1 and Ln fasting insulin, fasting blood glucose as categories (<100, 100–109 and ≥110) and Ln gamma‐glutamyl transferase or Ln alanine aminotransferase (mutually adjusted). Standard deviation for Ln(gamma‐glutamyl transferase) = 0.7; standard deviation for Ln(alanine aminotransferase) = 0.5. REF, reference; T, tertile.
Stratified analyses by drinking status, BMI and triglycerides
The incidence rates of type 2 diabetes mellitus in categories created by the combination of GGT and ALT were 6.1, 11.6 and 17.1 per 1,000 person‐years in “both low”, “either high” and “both high” categories, respectively (Table 4). The positive association remained significant in the fully‐adjusted model including fasting insulin and FBG. The significant associations were also observed in non‐ or light‐to‐moderate drinkers and in individuals with normal weight, suggesting that there are direct associations independent of these conditions associated with elevated GGT and ALT (P for difference in the associations according to drinking status and BMI >0.05). In addition, we observed significant associations of the combination of GGT and ALT with type 2 diabetes mellitus incidence in participants with a normal TC range, FBG <100 mg/dL and fasting insulin <10 μU (Table S1). However, the association might have been confounded by the presence of hypertriglyceridemia or low HDL‐c. Namely, the association of the combination of GGT and ALT appeared to be significant only in participants with hypertriglyceridemia or low HDL‐c.
Table 4.
Association of categories of gamma‐glutamyl transferase, alanine aminotransferase and their combination with type 2 diabetes mellitus incidence stratified by drinking status, obesity and triglycerides, Aichi cohort study 2002–2015, Japan
| Gamma‐glutamyl transferase (IU/L) | Alanine aminotransferase (IU/L) | Combination of gamma‐glutamyl transferase and alanine aminotransferase | |||||
|---|---|---|---|---|---|---|---|
| Low (T1 + T2) (8–48) | High (T3) (49–1,121) | Low (T1 + T2) (5–27) | High (T3) (28–294) | Both low | Either high | Both high | |
| All participants | |||||||
| (n = 1,812) | (n = 963) | (n = 1,825) | (n = 950) | (n = 1,417) | (n = 803) | (n = 555) | |
| No. incidents | 130 | 146 | 138 | 138 | 89 | 90 | 97 |
| Person‐years | 18,016 | 9,025 | 17,998 | 9,043 | 1,4174 | 7,665 | 5,201 |
| Crude incidence rate | 7.2 | 16.2 | 7.7 | 15.3 | 6.3 | 11.7 | 18.7 |
| HR 1 | REF | 1.9 (1.4–2.4) | REF | 1.6 (1.2–2.1) | REF | 1.6 (1.2–2.1) | 2.3 (1.7–3.2) |
| HR 2 | 1.4 (1.1–1.0) | 1.3 (1.0–1.7) | 1.4 (1.1–1.9) | 2.0 (1.5–2.8) | |||
| Drinking status | |||||||
| <20 g/day | (n = 1,444) | (n = 474) | (n = 1,284) | (n = 634) | (n = 1,107) | (n = 514) | (n = 297) |
| No. incidents | 108 | 75 | 85 | 98 | 69 | 55 | 59 |
| Person‐years | 14,475 | 4,432 | 12,890 | 6,018 | 11,218 | 4,928 | 2,761 |
| Crude incidence rate | 7.5 | 16.9 | 6.6 | 16.3 | 6.2 | 11.2 | 21.4 |
| HR 1 | REF | 1.9 (1.4 –2.5) | REF | 2.1 (1.5–2.9) | REF | 1.6 (1.1–2.3) | 2.8 (1.9–4.1) |
| HR 2 | 1.3 (0.9–1.8) | 1.7 (1.2–2.3) | 1.4 (0.9–2.1) | 2.4 (1.7–3.6) | |||
| ≥20 g/day | (n = 368) | (n = 489) | (n = 541) | (n = 316) | (n = 310) | (n = 289) | (n = 258) |
| No. incidents | 22 | 71 | 53 | 40 | 20 | 35 | 38 |
| Person‐years | 3,540 | 4,593 | 5,108 | 3,025 | 2,956 | 2,737 | 2,440 |
| Crude incidence rate | 6.2 | 15.5 | 10.4 | 13.2 | 6.8 | 12.8 | 15.6 |
| HR 1 | REF | 2.1 (1.2–3.4) | REF | 1.0 (0.6–1.5) | REF | 1.6 (0.9–2.8) | 1.7 (0.9–3.0) |
| HR 2 | 1.9 (1.1–3.3) | 0.7 (0.5–1.3) | 1.4 (0.8–2.4) | 1.6 (0.9–2.9) | |||
| Body mass index | |||||||
| <25 kg/m2 | (n = 1,485) | (n = 662) | (n = 1,539) | (n = 608) | (n = 1,216) | (n = 592) | (n = 339) |
| No. incidents | 93 | 80 | 104 | 69 | 69 | 59 | 45 |
| Person‐years | 14,886 | 6,375 | 15,332 | 5,929 | 12,257 | 5703 | 3,300 |
| Crude incidence rate | 6.2 | 12.5 | 6.8 | 11.6 | 5.63 | 10.3 | 13.6 |
| HR 1 | REF | 1.9 (1.3–2.6) | REF | 1.6 (1.2–2.2) | REF | 1.6 (1.2–2.3) | 2.2 (1.5–3.4) |
| HR 2 | 1.5 (1.1–2.2) | 1.3 (0.9–1.8) | 1.5 (1.0–2.2) | 2.0 (1.3–3.0) | |||
| ≥25 kg/m2 | (n = 327) | (n = 301) | (n = 286) | (n = 342) | (n = 201) | (n = 211) | (n = 216) |
| No. incidents | 37 | 66 | 34 | 69 | 20 | 31 | 52 |
| Person‐years | 3,130 | 2,650 | 2,666 | 3113.7 | 1,917 | 1,966 | 1,901 |
| Crude incidence rate | 11.8 | 24.9 | 12.8 | 22.2 | 10.4 | 15.8 | 27.4 |
| HR 1 | REF | 1.9 (1.2–2.9) | REF | 1.4 (1.0–2.0) | REF | 1.4 (0.8–2.5) | 2.3 (1.3–3.9) |
| HR 2 | 1.3 (0.8–2.1) | 1.2 (0.8–1.8) | 1.3 (0.7–2.3) | 2.0 (1.1–3.4) | |||
| Triglyceride | |||||||
| <150 mg/dL | (n = 1,398) | (n = 513) | (n = 1,397) | (n = 514) | (n = 1,134) | (n = 527) | (n = 250) |
| No. incidents | 90 | 51 | 91 | 50 | 64 | 53 | 24 |
| Person‐years | 13,894 | 4,941 | 13,876 | 4,959 | 11,374 | 5,022 | 2439 |
| Crude incidence rate | ,6.5 | 10.3 | 6.6 | 10.1 | 5.6 | 10.6 | 9.8 |
| HR 1 | REF | 1.4 (1.0–2.0) | REF | 1.4 (1.0–2.0) | REF | 1.7 (1.2–2.5) | 1.5 (0.9–2.5) |
| HR 2 | 1.2 (0.8–1.7) | 1.2 (0.8–1.8) | 1.5 (1.0–2.1) | 1.5 (0.9–2.4) | |||
| ≥150 mg/dL | (n = 414) | (n = 450) | (n = 428) | (n = 436) | (n = 283) | (n = 276) | (n = 305) |
| No. incidents | 40 | 95 | 47 | 88 | 25 | 37 | 73 |
| Person‐years | 4,122 | 4,084 | 4,122 | 4,084 | 2,800 | 2,643 | 2,763 |
| Crude incidence rate | 9.7 | 23.3 | 11.4 | 21.6 | 8.9 | 14.0 | 26.4 |
| HR 1 | REF | 2.4 (1.6–3.6) | REF | 1.8 (1.3–2.6) | REF | 1.5 (0.9–2.5) | 2.9 (1.8–4.6) |
| HR 2 | 1.7 (1.1–2.6) | 1.4 (0.9–2.1) | 1.4 (0.8–2.3) | 2.5 (1.5–4.0) | |||
HR 1: adjusted for age, family history of diabetes (yes/no), regular physical activity (yes/no), current, smoking status (yes/no), alcohol consumption (g/day), logarithmically transformed (Ln) body mass index (kg/m2), Ln triglycerides, Ln high‐density lipoprotein cholesterol and Ln total cholesterol. HR 2: adjusted for the same covariates as HR 1 and Ln fasting insulin, fasting blood glucose as categories (<100, 100–109 and ≥110), and Ln gamma‐glutamyl transferase or alanine aminotransferase (mutually adjusted), except relevant strata variables. REF, reference; T, tertile.
Incremental predictive value of adding GGT or/and ALT
The base model that included traditional risk factors for type 2 diabetes mellitus showed a C‐statistic of 0.747 (95% CI 0.717–0.778; Table 5). The addition of Ln‐GGT or/and Ln‐ALT to the base model increased the C‐statistic, and significantly improved IDI and NRI (P < 0.05). However, further adding GGT to the model (base + ALT) did not significantly improve IDI (P = 0.159) nor NRI (P = 0.106) anymore.
Table 5.
Incremental predictive value of gamma‐glutamyl transferase or/and alanine aminotransferase for type 2 diabetes mellitus incidence at 10 years, Aichi cohort study 2002–2015, Japan
| Prediction model | C‐statistic (95% CI) | IDI (95% CI) | P‐value | NRI (95% CI) | P‐value |
|---|---|---|---|---|---|
| Base model | 0.747 (0.717–0.778) | REF | <0.05 | REF | <0.05 |
| Base model + Ln‐GGT | 0.754 (0.724–0.784) | 0.007 (0.001–0.018) | 0.130 (0.055–0.198) | ||
| Base model | 0.747 (0.715–0.779) | REF | <0.05 | REF | <0.01 |
| Base model + Ln‐ALT | 0.754 (0.723–0.785) | 0.010 (0.002–0.022) | 0.130 (0.051–0.212) | ||
| Base model | 0.747 (0.714–0.780) | REF | <0.01 | REF | <0.01 |
| Base model + Ln‐GGT +Ln‐ ALT + Int | 0.755 (0.723–0.788) | 0.012 (0.006–0.026) | 0.134 (0.056–0.207) | ||
| Base model + Ln‐GGT | 0.754 (0.723–0.785) | REF | <0.05 | REF | 0.060 |
| Base model + Ln‐GGT + Ln‐ ALT + Int | 0.755 (0.725–0.786) | 0.004 (0.000–0.015) | 0.135 (–0.004–0.186) | ||
| Base model + Ln‐ALT | 0.754 (0.724–0.785) | REF | 0.159 | REF | 0.106 |
| Base model + Ln‐ALT + Ln‐ GGT + Int | 0.755 (0.725–0.785) | 0.002 (0.000–0.010) | 0.096 (−0.039–0.156) |
Total sample n = 2,775. Base model included age, family history of diabetes (yes/no), regular physical activity (yes/no), current, smoking status (yes/no), alcohol consumption (g/day), body mass index (kg/m2), triglycerides, high‐density lipoprotein cholesterol, total cholesterol, fasting insulin, fasting blood glucose as categories (<100, 100–109 and ≥110). ALT, alanine aminotransferase, CI, confidence interval; GGT, gamma‐glutamyl transferase; Int, interaction for logarithmically transformed gamma‐glutamyl transferase × logarithmically transformed alanine aminotransferase; IDI, integrated discrimination improvement; Ln, logarithmically transformed; NRI, net reclassification improvement; REF, reference; T, tertile.
Discussion
The present study showed that simultaneous elevation of GGT and ALT was significantly associated with the development of type 2 diabetes mellitus, even after adjustment for fasting insulin and FBG levels. Furthermore, the association was observed in non‐ or light‐to‐moderate drinkers or in participants with normal weight, totally excluding the possibility of confounding by these variables. Interestingly, the significant association was observed in participants with elevated triglyceride levels, but not in participants with normal triglyceride levels.
The present prospective study is consistent with and extends the findings of previous cross‐sectional studies15, 16. Adjustment for additional variables, such as fasting insulin and fasting blood glucose, was also carried out. Furthermore, stratified analyses were carried out by alcohol drinking status, overweight and hypertriglyceridemia in the present study. Significant associations observed in non‐ or light‐to‐moderate drinkers, as well as in individuals with normal weight, indicated that elevation of ALT and GGT would increase type 2 diabetes mellitus risk independent of alcohol intake and obesity. However, the association was observed only in participants with high triglyceride levels. This finding might indicate that the simultaneous presence of hepatic fat accumulation (high TG) and liver injury (high ALT), as well as oxidative stress (high GGT), is etiologically involved in the development of type 2 diabetes mellitus13, 14, 26, 27.
The implications of the present findings are threefold. First, participants with elevated ALT and GGT had approximately double the risk of type 2 diabetes mellitus incidence, independent of confounding variables, including alcohol intake, obesity, fasting insulin and FBG. Liver injury expressed by elevation of both GGT and ALT should be avoided to prevent type 2 diabetes mellitus. Second, investigations are warranted in those with elevated GGT and ALT for the cause of their elevation, so as to plan type 2 diabetes mellitus preventive measures. Finally, the molecular pathway leading from the concomitant elevation of triglycerides, ALT and GGT to type 2 diabetes mellitus development should be elucidated. We previously reported that the degree of intrahepatic fat accumulation was positively associated with blood levels of ALT independent of homeostasis model assessment for insulin resistance, which indicated that liver fat per se leads to liver damage28 and a state of increased oxidative stress3, 29, 30, 31.
The association between non‐alcoholic fatty liver disease and type 2 diabetes mellitus incidence could be explained by intrahepatic fat‐led systemic inflammation and insulin resistance14, 32, 33, 34. However, it remains to be elucidated whether early signs of liver damage by intrahepatic fat or non‐alcoholic fatty liver disease itself would lead to type 2 diabetes mellitus, or if insulin resistance would be responsible for intrahepatic fat accumulation35.
Regarding the incremental predictive value, a cross‐sectional study in Singapore and a prospective study in the Netherlands implied that the addition of liver enzymes improved the predictive value for type 2 diabetes mellitus15, 36. The present study extended the evidence to a long‐term prospective cohort study.
The strengths of the present study include the prospective cohort design with a 12‐year follow‐up period. Fasting insulin and FBG were included as a covariate, and stratified analyses were carried out by important variables. Furthermore, the predictive value of GGT and ALT were evaluated by reliable indicators, such as the C‐statistic, IDI and NRI. However, there were several limitations. First, it was not possible to differentiate non‐drinkers who might have quit drinking due to health problems. We included non‐drinking to be collapsed with light‐to‐moderate drinking in the present study. This might be a reason for the slightly higher incidence rate of type 2 diabetes mellitus in that group than in excessive drinking when GGT and/or ALT was high. Second, while creating categories for the combination of GGT and ALT, the category with only high ALT or GGT was collapsed, as type 2 diabetes mellitus incidence was not considered enough. Future studies are required to investigate the main effect of each liver enzyme and their joint effect on type 2 diabetes mellitus incidence. Third, women could not be evaluated because of the small sample size in the present study. Much larger studies are required to investigate the associations in women.
In conclusion, a concomitant increase of ALT and GGT was associated with a higher incidence of type 2 diabetes mellitus during a 12‐year follow‐up period in middle‐aged Japanese men, independent of alcohol drinking and weight. The positive association was not observed in men with normal triglyceride levels.
Disclosure
The authors declare no conflict of interest.
Supporting information
Table S1 Association of categories of gamma‐glutamyl transferase, alanine aminotransferase and their combination with type 2 diabetes mellitus incidence stratified by high‐density lipoprotein cholesterol, total cholesterol, fasting blood glucose and fasting insulin, Aichi cohort study 2002–2015, Japan.
Acknowledgments
The authors thank the participants and the healthcare personnel of the local government office. This work was supported in part by MEXT/JSPS KAKENHI (grant numbers 13470087 and 17390185 to Professor Emeritus Hideaki Toyoshima; 17790384, 22390133, 23659346, 26293153 and 18H03057 to Hiroshi Yatsuya; 16590499, 18590594, 20590641 and 23590787, 15K08802 to Koji Tamakoshi; and 25893088 and 16K19278 to Yuanying Li), Health and Labor Sciences research grants for Comprehensive Research on Cardiovascular and Life‐Style Related Diseases (H26‐Junkankitou [Seisaku]‐Ippan‐001 and H29‐Junkankitou [Seishuu]‐Ippan‐003) from the Ministry of Health, Labor and Welfare, and research grants from the Japan Atherosclerosis Prevention Fund (to Hiroshi Yatsuya), the Aichi Health Promotion Foundation (to Hiroshi Yatsuya) and the Uehara Memorial Fund (to Hiroshi Yatsuya).
J Diabetes Investig 2019; 10: 837–845
Research institution: Department of Public Health and Health Systems, Nagoya University Graduate School of Medicine, 65 Tsurumai‐cho, Showa‐ku, Nagoya 466‐8550, Japan.
References
- 1. NCD Risk Factor Collaboration . Worldwide trends in diabetes since 1980: a pooled analysis of 751 population‐based studies with 4.4 million participants. Lancet 2016; 387: 1513–1530. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Charvat H, Goto A, Goto M, et al Impact of population aging on trends in diabetes prevalence: a meta‐regression analysis of 160,000 Japanese adults. J Diabetes Investig 2015; 6: 533–542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Rui L. Energy metabolism in the liver. Compr Physiol 2014; 4: 177–197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Kunutsor SK, Apekey TA, Walley J. Liver aminotransferases and risk of incident type 2 Diabetes: a systematic review and meta‐analysis. Am J Epidemiol 2013; 178: 159–171. [DOI] [PubMed] [Google Scholar]
- 5. Fraser A, Harris R, Sattar N, et al Alanine aminotransferase, gamma‐glutamyltransferase, and incident diabetes: the British Women's Heart and Health Study and meta‐analysis. Diabetes Care 2009; 32: 741–750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Kunutsor SK, Abbasi A, Adler AI. Gamma‐glutamyl transferase and risk of type II diabetes: an updated systematic review and dose‐response meta‐analysis. Ann Epidemiol 2014; 24: 809–816. [DOI] [PubMed] [Google Scholar]
- 7. Abbasi A, Sahlqvist AS, Lotta L, et al A systematic review of biomarkers and risk of incident type 2 diabetes: an overview of epidemiological, prediction and aetiological research literature. PLoS ONE 2016; 11: e0163721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Nano J, Muka T, Ligthart S, et al Gamma‐glutamyltransferase levels, prediabetes and type 2 diabetes: a mendelian randomization study. Int J Epidemiol 2017; 46: 1400–1409. [DOI] [PubMed] [Google Scholar]
- 9. Noordam R, Smit RA, Postmus I, et al Assessment of causality between serum gamma‐glutamyltransferase and type 2 diabetes mellitus using publicly available data: a Mendelian randomization study. Int J Epidemiol 2016; 45: 1953–1960. [DOI] [PubMed] [Google Scholar]
- 10. Liu JX, Yeung SLA, Lin SL, et al Liver enzymes and risk of ischemic heart disease and type 2 diabetes mellitus: a Mendelian randomization study. Sci Rep 2016; 6: 38813. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Niemela O. Biomarkers in alcoholism. Clin Chim Acta 2007; 377: 39–49. [DOI] [PubMed] [Google Scholar]
- 12. Westerbacka J, Corner A, Tiikkainen M, et al Women and men have similar amounts of liver and intra‐abdominal fat, despite more subcutaneous fat in women: implications for sex differences in markers of cardiovascular risk. Diabetologia 2004; 47: 1360–1369. [DOI] [PubMed] [Google Scholar]
- 13. Lattuada G, Ragogna F, Perseghin G. Why does NAFLD predict type 2 diabetes? Curr DiabRep 2011; 11: 167–172. [DOI] [PubMed] [Google Scholar]
- 14. Byrne CD, Targher G. NAFLD: a multisystem disease. J Hepatol 2015; 62: S47–S64. [DOI] [PubMed] [Google Scholar]
- 15. Wang YL, Koh WP, Yuan JM, et al Association between liver enzymes and incident type 2 diabetes in Singapore Chinese men and women. BMJ Open Diabetes Res Care 2016; 4: e000296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Ko SH, Baeg MK, Han KD, et al Increased liver markers are associated with higher risk of type 2 diabetes. World J Gastroenterol 2015; 21: 7478–7487. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Wang C, Yatsuya H, Tamakoshi K, et al Association between parental history of diabetes and the incidence of type 2 diabetes mellitus differs according to the sex of the parent and offspring's body weight: a finding from a Japanese worksite‐based cohort study. Prev Med 2015; 81: 49–53. [DOI] [PubMed] [Google Scholar]
- 18. Tsugane S, Fahey MT, Sasaki S, et al Alcohol consumption and all‐cause and cancer mortality among middle‐aged Japanese men: seven‐year follow‐up of the JPHC study Cohort I. Japan Public Health Center. Am J Epidemiol 1999; 150: 1201–1207. [DOI] [PubMed] [Google Scholar]
- 19. WHO . Global Status Report on Noncommunicable Diseases 2014. Geneva, Switzerland: World Health Organization, 2014. [Google Scholar]
- 20. Cleeman JI, Grundy SM, Becker D, et al Executive summary of the third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel III). JAMA‐J Am Med Assoc 2001; 285: 2486–2497. [DOI] [PubMed] [Google Scholar]
- 21. WHO . Definition and diagnosis of diabetes mellitus and intermediate hyperglycemia :report of a WHO/IDF consultation. Geneva, Switzerland: World Health Organization, 2006. [Google Scholar]
- 22. Horst KW, Gilijamse PW, Koopman KE, et al Insulin resistance in obesity can be reliably identified from fasting plasma insulin. Int J Obes (Lond) 2015; 39: 1703–1709. [DOI] [PubMed] [Google Scholar]
- 23. Uno H, Cai TX, Pencina MJ, et al On the C‐statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med 2011; 30: 1105–1117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Yatsuya H. Risk prediction In: Vasan R, Sawyer D. (eds). The Encyclopedia of Cardiovascular Research and Medicine. Amsterdam, the Netherlands: Oxford Elsevier, 2018; 4: 315–318. [Google Scholar]
- 25. Pencina MJ, D'Agostino RB, Demler OV. Novel metrics for evaluating improvement in discrimination: net reclassification and integrated discrimination improvement for normal variables and nested models. Stat Med 2012; 31: 101–113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Ballestri S, Zona S, Targher G, et al Nonalcoholic fatty liver disease is associated with an almost twofold increased risk of incident type 2 diabetes and metabolic syndrome. Evidence from a systematic review and meta‐analysis. J Gastroenterol Hepatol 2016; 31: 936–944. [DOI] [PubMed] [Google Scholar]
- 27. Chen SC, Tsai SP, Jhao JY, et al Liver fat, hepatic enzymes, alkaline phosphatase and the risk of incident type 2 diabetes: a prospective study of 132,377 adults. Sci Rep 2017; 7: 4649. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Yatsuya H, Nihashi T, Li Y, et al Independent association of liver fat accumulation with insulin resistance. Obes Res Clin Pract 2014; 8: e350–e355. [DOI] [PubMed] [Google Scholar]
- 29. Sunny NE, Parks EJ, Browning JD, et al Excessive hepatic mitochondrial TCA cycle and gluconeogenesis in humans with nonalcoholic fatty liver disease. Cell Metab 2011; 14: 804–810. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Mari M, Caballero F, Colell A, et al Mitochondrial free cholesterol loading sensitizes to TNF‐ and Fas‐mediated steatohepatitis. Cell Metab 2006; 4: 185–198. [DOI] [PubMed] [Google Scholar]
- 31. Park JM, Lee JY, Lee DC, et al Serum ‐glutamyltransferase level and metabolic syndrome in children and adolescents: korean national health and nutrition examination survey. J Diabetes Investig 2018; 9: 522–528. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Kikuchi A, Takamura T. Where does liver fat go? A possible molecular link between fatty liver and diabetes. J Diabetes Investig 2017; 8: 152–154. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Adams LA, Anstee QM, Tilg H, et al Non‐alcoholic fatty liver disease and its relationship with cardiovascular disease and other extrahepatic diseases. Gut 2017; 66: 1138–1153. [DOI] [PubMed] [Google Scholar]
- 34. Furukawa Y, Tamura Y, Takeno K, et al Impaired peripheral insulin sensitivity in non‐obese Japanese patients with type 2 diabetes mellitus and fatty liver. J Diabetes Investig 2018; 9: 529–535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Matsuzaka T, Shimano H. Molecular mechanisms involved in hepatic steatosis and insulin resistance. J Diabetes Investig 2011; 2: 170–175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Abbasi A, Bakker SJL, Corpeleijn E, et al Liver function tests and risk prediction of incident type 2 diabetes: evaluation in two independent cohorts. PLoS ONE 2012; 7: 51496 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1 Association of categories of gamma‐glutamyl transferase, alanine aminotransferase and their combination with type 2 diabetes mellitus incidence stratified by high‐density lipoprotein cholesterol, total cholesterol, fasting blood glucose and fasting insulin, Aichi cohort study 2002–2015, Japan.