Abstract
Objective
We investigated the association of diabetes diagnosis and medication type with liver injury in individuals with clinical diabetes.
Methods
We analyzed 2,426 patients with clinical diabetes in the National Health and Nutrition Examination Survey conducted from 1999 to 2008. Elevated alanine transaminase (ALT) and aspartate transaminase (AST) levels were used as markers of liver injury. Participants were categorized into one of three categories: 1) clinical diabetes without physician’s diagnosis or diabetes medication; 2) diagnosed diabetes without diabetes medication; or 3) diagnosed diabetes with diabetes medication, further divided by classes of diabetes medications prescribed. We conducted logistic regression analysis to examine the relationship between diabetes diagnosis and medication type and elevated ALT and AST, adjusting for race/ethnicity, education, health insurance status, BMI category, alcohol consumption, physical activity, antihyperlipidemic agents, glycohemoglobin, C-reactive protein, viral hepatitis and liver disease.
Results
Participants with undiagnosed diabetes were more likely to have elevated ALT and AST levels (OR = 1.82, 95% CI 1.47, 2.42; OR = 1.99, 95% CI 1.46, 2.71, respectively). In contrast, there was no association between specific diabetes medication (i.e., sulfonylureas, biguanides/thiazolidinediones) and elevated ALT or AST levels among the treated. Our findings were confirmed in sensitivity analyses employing a lower threshold for ALT, and excluding individuals with viral hepatitis or liver disease.
Conclusion
We found that undiagnosed diabetes is associated with liver injury, compared to diagnosed diabetes with treatment. The effect of diabetes treatment on liver injury in individuals with diabetes remains uncertain.
Keywords: diabetes mellitus, alanine transaminase, aspartate transaminase, liver injury, biguanides
Introduction
Chronic liver disease and cirrhosis are leading causes of death due to digestive diseases in the U.S. (1) and entails considerable direct and indirect costs to patients and their family (1, 2). A common cause of chronic liver disease is non-alcoholic fatty liver disease (NAFLD), which is associated with higher risk of mortality, especially in the presence of cirrhosis and impaired glucose control (3). Risk factors for NAFLD include diabetes, obesity and hyperlipidemia (4). In individuals without diabetes, NAFLD and higher levels of alanine transminase and gamma-glutamyl transpeptidase, were found to predict incident diabetes and insulin resistance (5–9). Based on the association, it has been proposed that liver enzymes and NALFD may serve as useful biomarkers for diabetes (10).
Although no pharmacological agent has been approved specifically for treatment of NAFLD, some insulin-sensitizing agents for diabetic or pre-diabetic conditions show beneficial effects. For example, in the Diabetes Prevention Program, metformin treatment was found to lower alanine transaminase (ALT) levels, an indicator of hepatobiliary injury (11). Furthermore thiazolidinediones, another class of insulin sensitizing agent, was found to be associated with decreased hepatic steatosis and lower ALT levels (12). In addition, sustained reduction of ALT, independent of insulin resistance, was achieved in a trial of metformin and thiazolidinediones therapy (13). While the mechanism of the impact of insulin sensitizing agents is still unclear, it was hypothesized that part of the effect is mediated by weight loss, as demonstrated in the Diabetes Prevention Program (11). This is plausible, given that increased adiposity is associated with higher ALT levels (14, 15).
In light of the potential benefit of diabetes management on prevention of liver injury, we investigated the association between diabetes diagnosis status and liver injury, as measured by ALT and aspartate transaminase (AST), and how liver injury varies by type of diabetes medication in a representative sample of the U.S. population with clinical diabetes. We also conducted sub-group analyses to examine if the associations of diabetes diagnosis and medication and liver injury varied by overweight/obese status.
Methods
Study population
The National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC) has been conducting an ongoing series of cross-sectional surveys known as the National Health and Nutrition Examination Survey (NHANES) (16) to assess the health and nutritional status of adults and children in the United States. Recruited participants are interviewed and physically examined, and blood samples are collected for laboratory tests. For this study, we analyzed a sub-sample of the population surveyed between 1999 and 2008, who were 20 years of age or older and who met the clinical definition of diabetes at the time of the interview.
Diabetes
Following current guidelines set forth by the American Diabetes Association, diabetes was defined as a fasting blood glucose of ≥ 126 mg/dL, blood glucose level of ≥ 200 mg/dL by oral glucose tolerance testing, or a hemoglobin A1c (HbA1c) level of ≥ 6.5% (17). Because oral glucose tolerance test was not available for the years 1999–2003, we defined diabetes based on fasting blood glucose and HbA1c for these years. Plasma glucose levels were determined by the hexokinase enzymatic method at the Diabetes Diagnostic Laboratory at the University of Missouri-Columbia. HbA1c was measured also at the Diabetes Diagnostic Laboratory at the University of Missouri-Columbia using Primus CLC330 and Primus CLC 385 (Primus Corporation, Kansas City, MO) for years 1999–2004, Tosoh A1c 2.2 Plus Glycohemoglobin Analyzer for years 2005–2006 and Tosoh A1c G7 HPLC Glycohemoglobin Analyzer for years 2007–2008 (Tosoh Medics, San Francisco, CA). The administration of diabetes medication was determined through an in-home interview, which included an inventory of the participants’ medicine cabinet. Those with clinical diabetes were categorized into one of the following groups: 1) clinical diabetes without physician’s diagnosis of diabetes or diabetes medication, 2) diagnosed diabetes without diabetes medication, or 3) diagnosed diabetes treated with diabetes medication. Individuals with clinical diabetes who did not report physician’s diagnosis of diabetes but reported use of diabetes drugs were considered to have diagnosed diabetes. Participants treated for diabetes were further grouped by class of diabetes medications: 1) sulfonylureas only, 2) biguanides or thiazolidinediones only, 3) any combination of sulfonylureas, biguanides, thiazolidinediones, 4) other diabetes medications (e.g. alpha-glucosidase inhibitors) and 5) insulin.
Liver function enzymes
We considered two indicators of liver injury in the NHANES data: ALT and AST. From 2003 to 2008, both indicators were measured using a Beckman Synchron LX20 and Beckman UniCel DxC800 Synchron (Beckman Coulter, Brea, CA). For samples collected in 1999–2002 the Hitachi Model 704 multichannel analyzer (Boehringer Mannheim Diagnostics, Indianapolis, IN) was used. Following the NHANES III Laboratory Procedure, we defined elevated ALT levels as >40 U/L for males and >31 U/L for females, and elevated AST levels as >37 U/L for males and >31 U/L for females (18).
Covariates
We considered the following as potential confounders: age, sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic and other race), education level (11th grade or less, high school, some college or above), current health insurance status, HbA1c level, body mass index (BMI), C-reactive protein, hepatitis B virus (HBV) or hepatitis C virus (HCV) infection, alcohol intake (<12 drinks/year, ≥12 drinks/yr to 1 drink/day, > 1 drink/day), physical activity for leisure or for commute, use of anti-hyperlipidemic agents, and self-report of physician-diagnosed liver disease. C-reactive protein was measured by latex-enhanced nephelometry. Infection with HBV was defined as testing positive for antibody to HBV surface antigen by the AUSZYME Monoclonal Test (Abbott Diagnostics, Lake Forest, IL) and infection with HCV was defined as testing positive to HCV RNA by COBAS Amplicor HCV Monitor Test (Roche, Pleasanton, CA).
Statistical Analyses
Differences in the distribution of categorical covariates among diabetes diagnosis and treatment groups (diagnosed with medication, diagnosed without medication, undiagnosed) were assessed using chi-square tests, and pairwise differences in distributions of continuous covariates were assessed using the Wilcoxon rank-sum test. Differences in ALT and AST distribution by diabetes medication type were examined using Kruskal Wallis analysis of variance. The bivariate associations of diabetes diagnosis and medication category and other variables with the outcomes elevated ALT and elevated AST status were assessed using logistic regression unadjusted for other variables. To assess for confounder-independent association between diabetes diagnosis and medication category and liver injury, we conducted multivariate logistic regression adjusting for all potential confounders listed above. We performed two sets of logistic regression analyses, comparing diagnosed diabetes without medication and undiagnosed diabetes to diagnosed diabetes with medication as the referent group, and separately comparing diabetes treated with non-insulin diabetes drugs (i.e. sulfonylureas, insulin sensitizers) to insulin-treated diabetes as the referent group. We also conducted stratified analyses by BMI categories of <25 kg/m2, 25–30 kg/m2 and ≥30kg/m2. We performed sensitivity analyses for ALT using a threshold of >29 U/L for males and >22 U/L for females following a recent finding that these cutoffs yield maximum correct classification of individuals at high risk for liver disease (19). We also performed sensitivity analysis excluding individuals with HBV/HCV infection and those with liver disease, who are expected to have considerable liver injury. Significance of associations was interpreted at P-value of less than 0.05. All statistical analyses were conducted with SAS version 9.2 (SAS Institute, Cary, NC).
Results
A total of 51,623 individuals were interviewed for NHANES between years 1999 and 2008. Of these, 26,246 were 20 years of age or older, 2,758 had diabetes and 2,426 had complete data on liver enzymes and other variables used in analyses; the latter group constituted the analytic sample. The age and sex distribution of individuals excluded due to missing data was not different from that of individuals included in the analyses.
Among participants with glucose or glycohemoglobin levels indicative of diabetes, 818 were undiagnosed, 251 were diagnosed but were not taking any diabetes medications, and 1,357 were taking at least 1 diabetes medication. Median ALT and AST levels were higher in those with undiagnosed diabetes compared to participants with diagnosed diabetes, either with or without medication (P-values < 0.0001, Table 1). Median ALT and AST levels were similar between people with diagnosed diabetes with and without medication. Those with undiagnosed diabetes or diabetes without medication were younger, more likely to be male, less likely to be insured, drank more alcohol and were less likely to be on anti-hyperlipidemic agents compared to those with diabetes with medication (Table 1). Among those on anti-diabetic medication, 267 (20%) were taking a sulfonylurea agent only, 305 (21%) were taking a biguanide agent or thiazolidinedione agent only, 341 (27%) were taking a combination of sulfonylurea, biguanide and/or thiazolidinedione agents, 91 (7%) were taking other diabetes drugs and 353 (26%) were on insulin. Median ALT and AST levels did not vary by diabetes treatment type (P=0.13 for ALT, p=0.73 for AST by Kruskal Wallis analysis of variance).
Table 1.
Sample characteristics of individuals with clinical diabetes mellitus by diabetes diagnosis and treatment status, NHANES 1999–2008
| Variable | Treated DM | Diagnosed DM not treated |
P-value (vs. treated) |
Undiagnosed DM |
P-value (vs. untreated diagnose d DM) |
P-value (vs. treated DM) |
|---|---|---|---|---|---|---|
| N (%) or median (IQR) |
N (%) or median (IQR) |
N (%) or median (IQR) |
||||
| Total N | 1357 | 251 | 818 | |||
| Liver enzyme levels | ||||||
| Alanine transaminase (U/L) | 21 (17, 28) | 22 (17, 29) | 0.45 | 25 (19, 36) | <0.0001 | <0.0001 |
| Aspartate transaminase (U/L) | 22 (18, 27) | 22 (18, 28) | 0.95 | 24 (20, 31) | <0.0001 | <0.0001 |
| Age group (yrs) | 0.15 | 0.42 | 0.003 | |||
| 20–39 | 73 (5.4%) | 17 (6.8%) | 72 (8.8%) | |||
| 40–59 | 392 (29%) | 85 (34%) | 249 (30%) | |||
| 60+ | 892 (66%) | 149 (59%) | 497 (61%) | |||
| Male sex | 670 (49%) | 138 (55%) | 0.10 | 468 (57%) | 0.53 | 0.0004 |
| Race / ethnicity | 0.62 | 0.008 | 0.0002 | |||
| Non-Hispanic Black | 371 (27%) | 67 (27%) | 170 (21%) | |||
| Hispanic | 436 (32%) | 89 (35%) | 234 (30%) | |||
| Other | 39 (2.9%) | 9 (2.6%) | 25 (3.1%) | |||
| Non-Hispanic White | 511 (38%) | 86 (34%) | 380 (46%) | |||
| Years of school completed | 0.08 | 0.003 | 0.09 | |||
| 11th grade or less | 617 (45%) | 132 (53%) | 335 (41%) | |||
| High school / GED | 299 (22%) | 43 (17%) | 205 (25%) | |||
| Some college or above | 441 (33%) | 76 (30%) | 278 (34%) | |||
| Current health insurance | 1207 (89%) | 196 (78%) | <0.0001 | 554 (81%) | 0.26 | <0.0001 |
| Body mass index | 0.003 | 0.02 | 0.79 | |||
| <25.0 kg/m2 | 188 (14%) | 52 (21%) | 117 (14%) | |||
| 25.0–29.9 kg/m2 | 427 (31%) | 87 (35%) | 266 (33%) | |||
| ≥30.0 kg/m2 | 742 (55%) | 112 (45%) | 435 (53%) | |||
| Alcohol intake | 0.03 | 0.02 | <0.0001 | |||
| 0 – 12 drinks in the past year | 577 (43%) | 90 (36%) | 237 (29%) | |||
| >12 drinks/year to ≤1 drink/day | 478 (35%) | 94 (37%) | 294 (36%) | |||
| 2–4 drinks/day | 231 (17%) | 43 (17%) | 215 (26%) | |||
| > 4 drinks / day | 71 (5.2%) | 24 (9.6%) | 72 (8.8%) | |||
| Leisure activity in the past 30 days | 0.11 | 0.39 | 0.47 | |||
| None | 809 (60%) | 154 (61%) | 502 (61%) | |||
| Moderate | 391 (29%) | 59 (24%) | 216 (26%) | |||
| Vigorous | 157 (12%) | 38 (15%) | 100 (12%) | |||
| Commute by walking or biking | 238 (18%) | 45 (18%) | 0.88 | 168 (21%) | 0.37 | 0.08 |
| Use of antihyperlipidemic agents | 628 (46%) | 61 (24%) | <0.0001 | 192 (23%) | 0.79 | <0.0001 |
| Glycohemoglobin (%) | 7.5 (6.9, 8.7) | 7.4 (6.7, 9.2) | 0.005 | 6.5 (5.9, 7.1) | <0.0001 | <0.0001 |
| <7.0% | 380 (28%) | 87 (35%) | 464 (70%) | |||
| 7–8.9% | 679 (50%) | 98 (39%) | 155 (19%) | |||
| ≥9.0% | 295 (22%) | 66 (26%) | 76 (11%) | |||
| C-reactive protein (mg/L) | 3.4 (1.5, 7.6) | 3.1 (1.3, 8.4) | 0.06 | 4.1 (1.8, 8.4) | 0.0006 | 0.01 |
| HBV or HCV infection | 32 (2.4%) | 4 (1.6%) | 0.45 | 14 (1.7%) | 0.90 | 0.31 |
| Liver disease | 73 (5.4%) | 13 (5.2%) | 0.90 | 44 (5.4%) | 0.90 | 1.00 |
DM = diabetes mellitus; IQR = interquartile range; HBV = hepatitis B virus, HCV = hepatitis C virus
P-values obtained by chi-square tests (categorical variables) or Wilcoxon rank-sum test (continuous variables).
ALT and AST levels were elevated in 407 (17%) and 312 (13%) of the analytic sample, respectively. In bivariate analysis (Table 2), those with undiagnosed diabetes were more likely to have elevated ALT levels (OR = 1.74, 95% CI 1.39, 2.18) and AST levels (OR = 1.75, 95% CI 1.37, 2.25) compared to those who were taking diabetes medications. Proportions of elevated ALT and AST were not different in those who were treated with insulin-sensitizers (biguanide or a thiazolidinedione) compared to those treated with insulin. Younger individuals, Hispanics, those with higher alcohol intake, and those with HBV/HCV infection or liver disease were more likely to have elevated ALT levels, while insured individuals and those taking antihyperlipidemic agents were less likely to have elevated ALT levels. Associations with elevated AST levels yielded similar results; younger individuals, those who commute by walking or biking, those with higher alcohol intake, and those with HBV/HCV infection or liver disease were more likely to have elevated AST levels, while insured individuals and those taking antihyperlipidemic agents were less likely to have elevated AST levels. Higher BMI was associated with both elevated ALT and elevated AST levels (Table 2).
Table 2.
Bivariate logistic regression results for the outcomes ALT elevation and AST elevation among individuals with clinical diabetes, NHANES 1999–2008
| Explanatory variable | Elevated ALT (N=407) n (%) or median (IQR) |
Not elevated ALT (N=2019) n (%) or median (IQR) |
Unadjusted OR for elevated ALT (95% CI) |
p-value | Elevated AST (N=312) n (%) or median (IQR) |
Not elevated AST (N=2114) n (%) or median (IQR) |
Unadjusted OR for elevated AST (95% CI) |
p-value |
|---|---|---|---|---|---|---|---|---|
| Analysis of all individuals (N = 2,426) | ||||||||
| Diabetes category | ||||||||
| Undiagnosed diabetes | 181 (44%) | 637 (32%) | 1.74 (1.39, 2.18) | <0.0001 | 143 (46%) | 675 (32%) | 1.75 (1.37, 2.25) | <0.0001 |
| Diagnosed untreated diabetes | 37 (9.1%) | 214 (11%) | 1.07 (0.73, 1.56) | 0.74 | 24 (7.7%) | 227 (11%) | 0.88 (0.56, 1.39) | 0.59 |
| Treated diabetes | 189 (46%) | 1168 (58%) | Ref | 145 (46%) | 1212 (57%) | Ref | ||
| Age group (yrs) | ||||||||
| 20–39 | 50 (12%) | 112 (5.6%) | 3.56 (2.46, 5.15) | <0.0001 | 36 (12%) | 126 (6.0%) | 2.51 (1.67, 3.76) | <0.0001 |
| 40–59 | 185 (45%) | 541 (27%) | 2.73 (2.16, 3.43) | <0.0001 | 118 (38%) | 608 (29%) | 1.70 (1.32, 2.20) | <0.0001 |
| 60+ | 172 (42%) | 1366 (68%) | Ref | 158 (51%) | 1380 (65%) | Ref | ||
| Sex | ||||||||
| Male | 216 (53%) | 1060 (53%) | 1.02 (0.82, 1.26) | 0.83 | 148 (47%) | 1128 (53%) | 0.78 (0.62, 0.99) | 0.044 |
| Female | 20191 (47%) | 959 (48%) | Ref | 164 (53%) | 986 (47%) | Ref | ||
| Race / ethnicity | ||||||||
| Non-Hispanic Black | 61 (15%) | 547 (27%) | 0.55 (0.40, 0.75) | 0.0002 | 53 (17%) | 555 (26%) | 0.63 (0.45, 0.89) | 0.007 |
| Hispanic | 167 (41%) | 601 (30%) | 1.38 (1.08, 1.75) | 0.009 | 120 (38%) | 648 (31%) | 1.23 (0.94, 1.61) | 0.13 |
| Other | 15 (3.7%) | 58 (2.9%) | 1.28 (0.71, 2.32) | 0.41 | 11 (3.5%) | 62 (2.9%) | 1.18 (0.60, 2.29) | 0.63 |
| Non-Hispanic White | 164 (40%) | 813 (40%) | Ref | 128 (41%) | 849 (40%) | Ref | ||
| Years of school completed | ||||||||
| 11th grade or less | 911 (45%) | 173 (43%) | 0.89 (0.70, 1.14) | 0.37 | 141 (45%) | 943 (45%) | 1.11 (0.84, 1.47) | 0.47 |
| High school / GED | 452 (22%) | 95 (23%) | 0.99 (0.75, 1.33) | 0.97 | 77 (25%) | 470 (22%) | 1.22 (0.89, 1.69) | 0.22 |
| Some college or above | 2656 (32%) | 139 (34%) | Ref | 94 (30%) | 701 (33%) | Ref | ||
| Covered by health insurance at the time of survey | ||||||||
| Insured | 306 (75%) | 1762 (87%) | 0.44 (0.34, 0.57) | <0.0001 | 241 (77%) | 1827 (86%) | 0.53 (0.40, 0.71) | <0.0001 |
| Uninsured | 101 (25%) | 257 (13%) | Ref | 71 (23%) | 287 (14%) | Ref | ||
| Body mass index (kg/m2) | 32 (28, 37) | 30 (27, 35) | 1.04 (1.02, 1.05) | <0.0001 | 31 (27, 37) | 30 (27, 35) | 1.02 (1.00, 1.04) | 0.02 |
| <25.0 kg/m2 | 36 (8.9%) | 321 (16%) | Ref | 36 (12%) | 321 (15%) | Ref | ||
| 25.0 – 29.9 kg/m2 | 112 (28%) | 668 (33%) | 1.49 (1.00, 2.22) | 0.05 | 93 (30%) | 687 (33%) | 1.20 (0.80, 1.81) | 0.37 |
| >=30.0 kg/m2 | 259 (64%) | 1030 (51%) | 2.23 (1.54, 3.23) | <0.0001 | 183 (59%) | 1106 (52%) | 1.47 (1.00, 2.14) | 0.048 |
| Alcohol intake | ||||||||
| 0 – 12 drinks in the past year | 129 (32%) | 775 (38%) | Ref | 97 (31%) | 807 (38%) | Ref | ||
| >12 drinks/year to ≤1 drink/day | 118 (29%) | 748 (37%) | 0.94 (0.72, 1.23) | 0.64 | 107 (34%) | 759 (36%) | 1.16 (0.87, 1.55) | 0.32 |
| 2–4 drinks/day | 111 (27%) | 378 (19%) | 1.76 (1.33, 2.34) | <0.0001 | 73 (23%) | 416 (20%) | 1.46 (1.05, 2.02) | 0.02 |
| > 4 drinks / day | 49 (12%) | 118 (5.8%) | 2.49 (1.70, 3.65) | <0.0001 | 35 (11%) | 132 (6.2%) | 2.20 (1.43, 3.38) | 0.0003 |
| Leisure activity in the past 30 days | ||||||||
| None | 236 (58%) | 1229 (61%) | Ref | 198 (63%) | 1267 (60%) | Ref | ||
| Moderate | 115 (28%) | 551 (27%) | 1.09 (0.85, 1.39) | 0.50 | 76 (24%) | 590 (28%) | 0.83 (0.62, 1.10) | 0.19 |
| Vigorous | 56 (14%) | 239 (12%) | 1.22 (0.89, 1.69) | 0.22 | 38 (12%) | 257 (12%) | 0.95 (0.65, 1.38) | 0.78 |
| Commute | ||||||||
| Walking or biking | 87 (21%) | 1655 (82%) | 1.24 (0.95, 1.61) | 0.11 | 241 (77%) | 1735 (82%) | 1.38 (1.04, 1.83) | 0.03 |
| Other | 320 (79%) | 264 (18%) | Ref | 72 (23%) | 379 (18%) | Ref | ||
| Use of antihyperlipidemic agents | ||||||||
| Yes | 114 (28%) | 1252 (62%) | 0.63 (0.50, 0.80) | 0.0001 | 215 (69%) | 1330 (63%) | 0.76 (0.59, 0.98) | 0.04 |
| No | 23 293 (72%) | 767 (38%) | Ref | 97 (31%) | 784 (38%) | Ref | ||
| Glycohemoglobin (%) | ||||||||
| <7.0% | 159 (39%) | 883 (44%) | Ref | 128 (41%) | 914 (43%) | Ref | ||
| 7–8.9% | 158 (39%) | 774 (38%) | 1.13 (0.89, 1.44) | 0.31 | 119 (38%) | 813 (39%) | 1.05 (0.80, 1.37) | 0.75 |
| >=9.0% | 89 (22%) | 359 (18%) | 1.38 (1.03, 1.84) | 0.029 | 64 (21%) | 384 (18%) | 1.19 (0.86, 1.64) | 0.29 |
| C-reactive protein (mg/L) | 4.1 (1.9, 8.6) | 3.5 (1.5, 7.6) | 1.006 (1.000, 1.013) | 0.07 | 4.3 (1.6, 9.2) | 3.5 (1.5, 7.6) | 0.992 (0.978, 1.006) | 0.43 |
| HBV or HCV infection | ||||||||
| Infected | 28 (6.9%) | 22 (1.1%) | 6.71 (3.80, 11.9) | <0.0001 | 27 (8.7%) | 23 (1.1%) | 8.63 (4.88, 15.3) | <0.0001 |
| Uninfected | 384 (94%) | 2002 (99%) | Ref | 290 (93%) | 2096 (99%) | Ref | ||
| Liver disease | ||||||||
| Liver disease | 63 (15%) | 67 (3.3%) | 5.34 (3.72, 7.68) | <0.0001 | 58 (19%) | 72 (3.4%) | 6.49 (4.49, 9.40) | <0.0001 |
| No liver disease | 344 (85%) | 1952 (97%) | Ref | 254 (81%) | 2042 (97%) | Ref | ||
| Analysis of individuals with treated diabetes (N = 1,357) | ||||||||
| Treatment groups | ||||||||
| Sulfonylureas | 36 (19%) | 231 (20%) | 0.99 (0.62, 1.58) | 0.97 | 22 (15%) | 245 (20%) | 0.77 (0.44, 1.33) | 0.35 |
| Biguanide/Thiazolidinediones | 40 (19%) | 265 (20%) | 0.96 (0.61, 1.51) | 0.86 | 31 (21%) | 273 (22%) | 0.97 (0.58, 1.60) | 0.89 |
| Combination | 51 (27%) | 290 (25%) | 1.12 (0.73, 1.71) | 0.61 | 45 (31%) | 296 (24%) | 1.30 (0.82, 2.06) | 0.27 |
| Other drug | 14 (7.4%) | 77 (6.6%) | NE | NE | 10 (6.9%) | 81 (6.7%) | NE | NE |
| Insulin | 48 (25%) | 305 (26%) | Ref | 37 (26%) | 316 (26%) | Ref |
CI = confidence interval, HBV = hepatitis B virus, HCV = hepatitis C virus, NE = not estimated due to low sample size
When adjusted for potential confounders, the associations between undiagnosed diabetes and elevated ALT and AST levels remained significant (OR = 1.82, 95% CI 1.47, 2.42 for ALT, OR = 1.99, 95% CI 1.46, 2.71 for AST) (Table 3). Adjusting for all potential confounders considered, those treated with insulin sensitizers (biguanides and/or thiazolidinediones), sulfonylureas, or a combination of these frequently-prescribed diabetes drugs were not more or less likely to have higher levels of ALT or AST compared to insulin-treated individuals. (Table 3) Multivariate analysis was not conducted for those treated with other diabetes drugs due to small sample size. Sensitivity analysis using a lower cut-off for ALT levels confirmed our finding for undiagnosed diabetes (OR = 1.98, 95% CI 1.59, 2.46), as well as the null findings for insulin-sensitizers and sulfonylureas. A second sensitivity analysis excluding HBV or HCV infected individuals and those with liver disease also confirmed our findings for association of undiagnosed diabetes with elevated ALT levels (OR = 2.11, 95% 1.56, 2.84) and with elevated AST levels (OR = 2.49, 95% CI 1.78, 3.48).
Table 3.
Results of multivariate logistic regression analyses for the outcomes ALT elevation and AST elevation among individuals with clinical diabetes, NHANES 1999–2008
| Variable | Adjusted OR for elevated AST |
p-value | Adjusted OR for elevated AST |
p-value |
|---|---|---|---|---|
| Analysis of all individuals (N = 2,426) | ||||
| Diabetes category | ||||
| Undiagnosed diabetes | 1.82 (1.37, 2.42) | <0.0001 | 1.99 (1.46, 2.71) | 0.0001 |
| Diagnosed untreated diabetes | 1.04 (0.69, 1.59) | 0.84 | 0.89 (0.55, 1.45) | 0.64 |
| Treated diabetes | Ref | Ref | ||
| Analysis of individuals with treated diabetes (N = 1,357) | ||||
| Diabetes medication type | ||||
| Sulfonylureas | 1.12 (0.66, 1.90) | 0.68 | 0.80 (0.43, 1.51) | 0.50 |
| Biguanide/Thiazolidinediones | 1.05 (0.64, 1.73) | 0.84 | 1.12 (0.64, 1.95) | 0.69 |
| Combination | 1.08 (0.67, 1.74) | 0.84 | 1.32 (0.79, 2.22) | 0.29 |
| Insulin | Ref | Ref |
Each analysis was adjusted for all variables listed in Table 2.
When we conducted stratified analyses by BMI categories of <25, 25–29, ≥30 kg/m2, corresponding to normal weight, overweight and obese, we found a similar association for elevated ALT and AST levels with undiagnosed diabetes among obese individuals (OR = 1.83, 95% CI 1.28, 2.61 for ALT, OR = 1.87, 95%CI 1.25, 2.80 for AST) (Table 4). While the associations were not statistically significant for individuals in the normal and overweight categories, there were trends towards positive associations and a test for heterogeneity of the effect by BMI level was insignificant (p=0.86 for ALT, p=0.99 for AST)
Table 4.
Results of multivariate logistic regression analyses for the outcomes ALT elevation and AST elevation among individuals with clinical diabetes, NHANES 1999–2008, stratified by BMI categories
| Variable | Adjusted OR for elevated ALT |
p-value | Adjusted OR for elevated AST |
p-value |
|---|---|---|---|---|
| Analysis of normal weight (n=357) | ||||
| Diabetes category | ||||
| Undiagnosed diabetes | 2.40 (0.86, 6.68) | 0.095 | 2.27 (0.81, 6.41) | 0.12 |
| Diagnosed untreated diabetes | 1.33 (0.42, 4.19) | 0.63 | 0.17 (0.02, 1.49) | 0.11 |
| Treated diabetes | Ref | Ref | ||
| Analysis of overweight (n=780) | ||||
| Diabetes category | ||||
| Undiagnosed diabetes | 1.50 (0.88, 2.55) | 0.13 | 1.89 (1.09, 2.30) | 0.02 |
| Diagnosed untreated diabetes | 1.08 (0.51, 2.27) | 0.84 | 0.81 (0.33, 2.02) | 0.66 |
| Treated diabetes | Ref | Ref | ||
| Analysis of obese (n=1289) | ||||
| Diabetes category | ||||
| Undiagnosed diabetes | 1.83 (1.28, 2.61) | 0.001 | 1.87 (1.25, 2.80) | 0.002 |
| Diagnosed untreated diabetes | 1.01 (0.56, 1.81) | 0.98 | 1.28 (0.67, 2.43) | 0.45 |
| Treated diabetes | Ref | Ref |
Each analysis was adjusted for all variables listed in Table 2.
Discussion
In this study of liver injury among individuals with clinical diabetes in a nationally representative sample of U.S. population, we found that those with undiagnosed diabetes were more likely to have elevated levels of ALT and AST compared to those who were taking diabetes medication. Elevated ALT and AST levels are strong laboratory-based indications of NAFLD (4). Insulin resistance in individuals with undiagnosed diabetes could drive accumulation of lipids in hepatocytes, causing NALFD (4), which increases vulnerability to cirrhosis and resulting mortality (3). The association may also be driven by the effect of liver injury or NAFLD on incident diabetes, which has been reported in previous observational studies (5–9). We were unable to attribute the association to the lack of medication for diabetes, as those with diagnosed diabetes without medication exhibited similar levels of ALT and AST as those treated with diabetes medication. Associations with ALT and AST levels were not attenuated after adjustment for other factors associated with liver function, including alcohol consumption, insurance status, BMI, HBV or HCV infection and liver disease, or factors related to diabetes, such as physical activity and glycohemoglobin. This suggests that variables other than these known risk factors of liver disease influence liver enzyme levels in those with undiagnosed diabetes. Possible contributors may include continuity of care that would influence detection and management of diabetes (20, 21) and related liver conditions. People with undiagnosed diabetes may have higher allostatic load, or physiological responses to environmental stress, which is linked to accumulation of fat, inflammation, and oxidative stress that can negatively impact liver health (22).
Regarding the effect of diabetes medications on liver enzyme levels, our findings stand in contrast to that of previous studies that found that treatment with biguanides or thiazolidinediones was associated with reduced ALT levels among individuals at risk for diabetes (11–13). Variability in results may arise from different methodologies or study populations. Our study was a cross-sectional study of a representative sample of people with diabetes in the U.S. while other studies were of randomized controlled design recruiting more homogenous group of participants with pre-diabetic conditions or hepatic steatosis (11–13). While our study is more vulnerable to confounding bias, it is possible that the effect on ALT is less pronounced in those who already have diabetes, while more apparent in those who are en route to developing diabetes as examined in previous trials (11, 13). As demonstrated in this study, those with undiagnosed diabetes appeared particularly at risk for elevated levels of ALT compared to those who were already diagnosed, despite lower levels of glycohemoglobin in the undiagnosed. The discordant results may also be due to differences in the referent groups. The referent group in our study sample of people with diabetes consisted of insulin treated individuals and diabetic regimen was not randomly assigned. Previous studies on insulin sensitizers and hepatic injury were conducted in people without diabetes and comparison groups were randomly assigned to either no treatment or placebo (11, 13) Thus, the non-random distribution of diabetic regimens may have introduced unmeasured confounding in our study. However, randomized placebo-controlled trials are not feasible in people with diabetes, who are recommended to take oral medications or insulin by consensus guidelines. Given that insulin resistance is correlated with higher ALT levels (23), we expected that people who took insulin-sensitizers would show more favorable ALT levels compared to those who took insulin or sulfonylureas that increase systemic levels of insulin. However, there was no clear difference in liver enzyme levels across all classifications of diabetes drugs. It is possible that the effect of the hypoglycemic agents were diluted due to non-adherence to treatment regimen.
Taking the findings from our study and together with those from other studies demonstrating higher levels of ALT in undiagnosed diabetes compared to those with impaired fasting glucose or normoglycemic individuals (15, 24), we conclude that in the continuum of progression from no diabetes to insulin-dependent diabetes, vulnerability to liver injury is highest in those with undiagnosed diabetes.
Our study was limited in that it employed cross-sectional data, and therefore could not examine the longitudinal effect of diabetes diagnosis and treatment on liver injury. Prospective studies with repeated measures of liver function enzymes would be particularly informative considering that effects of diabetes treatment take time to manifest and that liver enzymes are subject to temporal variation, especially in the presence of co-morbid conditions that require hemodialysis (25). In addition, the number of people taking each type of anti-diabetic medication may not have been large enough to detect nuanced differences in prevalence of elevated liver enzyme levels. There was also sparse data on specific liver disease conditions, such as non-alcoholic steatohepatitis or fatty liver disease that would better serve as definitive markers of liver injury. The study lacked detailed data on healthcare utilization and environmental stress that could have been correlated with diabetes diagnosis status and linked to liver health and may have thus confounded the relationship between undiagnosed diabetes and liver injury. It is also possible that those with diagnosed diabetes could have received better clinical follow-up with treatment for reversible causes that has positive effects on the liver. Finally, part of the NHANES data (1999–2003) lacked information on oral glucose tolerance testing, through which we may have missed those with true diabetes, but did not meet the definition of diabetes by fasting blood glucose or hemoglobin A1c. This decreased our sample size of people with diabetes, limiting the power of the study.
Despite the limitations, our study has a number of strengths. We employed a dataset representative of non-institutionalized people residing in the U.S. and thus our findings are generalizable to the US adult population living with diabetes. Also, we were able to consider multiple potential confounders relevant to liver health, including alcohol consumption and hepatitis B and C infection. Furthermore, systematic data on fasting blood glucose and glycohemoglobin permitted us to identify those who had undiagnosed diabetes.
In sum, we found that undiagnosed diabetes is associated with liver injury, as indicated by elevated levels of ALT and AST, compared to diagnosed diabetes with medication. Treatment with insulin-sensitizers was not associated with better liver enzyme levels compared to insulin treatment. Our findings suggest that early detection and treatment of diabetes might improve liver health, while further studies are needed to compare the effects of different diabetes medications on liver injury.
Acknowledgment
This study was supported by NIH NCI R25 Training Grant for career development in cancer prevention and control research.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Author contributions
CYJ was responsible for conception of the study, the analysis, and the first draft of the manuscript. CYJ, CKR, CMC, ZFZ contributed to interpretation of the results and critical revision of the manuscript.
References
- 1.Sandler RS, Everhart JE, Donowitz M, Adams E, Cronin K, Goodman C, et al. The burden of selected digestive diseases in the United States. Gastroenterology. 2002;122(5):1500–1511. doi: 10.1053/gast.2002.32978. [DOI] [PubMed] [Google Scholar]
- 2.Rakoski MO, McCammon RJ, Piette JD, Iwashyna TJ, Marrero JA, Lok AS, et al. Burden of cirrhosis on older Americans and their families: analysis of the health and retirement study. Hepatology. 2012;55(1):184–191. doi: 10.1002/hep.24616. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Adams LA, Lymp JF, St Sauver J, Sanderson SO, Lindor KD, Feldstein A, et al. The natural history of nonalcoholic fatty liver disease: a population-based cohort study. Gastroenterology. 2005;129(1):113–121. doi: 10.1053/j.gastro.2005.04.014. [DOI] [PubMed] [Google Scholar]
- 4.Angulo P. Nonalcoholic fatty liver disease. N Engl J Med. 2002;346(16):1221–1231. doi: 10.1056/NEJMra011775. [DOI] [PubMed] [Google Scholar]
- 5.Monami M, Bardini G, Lamanna C, Pala L, Cresci B, Francesconi P, et al. Liver enzymes and risk of diabetes and cardiovascular disease: results of the Firenze Bagno a Ripoli (FIBAR) study. Metabolism. 2008;57(3):387–392. doi: 10.1016/j.metabol.2007.10.015. [DOI] [PubMed] [Google Scholar]
- 6.Nguyen QM, Srinivasan SR, Xu JH, Chen W, Hassig S, Rice J, et al. Elevated liver function enzymes are related to the development of prediabetes and type 2 diabetes in younger adults: the Bogalusa Heart Study. Diabetes Care. 2011;34(12):2603–2607. doi: 10.2337/dc11-0919. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Choi JH, Rhee EJ, Bae JC, Park SE, Park CY, Cho YK, et al. Increased Risk of Type 2 Diabetes in Subjects with Both Elevated Liver Enzymes and Ultrasonographically Diagnosed Nonalcoholic Fatty Liver Disease: A 4-year Longitudinal Study. Arch Med Res. 2013 doi: 10.1016/j.arcmed.2013.01.007. [DOI] [PubMed] [Google Scholar]
- 8.Zhang Y, Lu X, Hong J, Chao M, Gu W, Wang W, et al. Positive correlations of liver enzymes with metabolic syndrome including insulin resistance in newly diagnosed type 2 diabetes mellitus. Endocrine. 2010;38(2):181–187. doi: 10.1007/s12020-010-9369-6. [DOI] [PubMed] [Google Scholar]
- 9.Bonnet F, Ducluzeau PH, Gastaldelli A, Laville M, Anderwald CH, Konrad T, et al. Liver enzymes are associated with hepatic insulin resistance, insulin secretion, and glucagon concentration in healthy men and women. Diabetes. 2011;60(6):1660–1667. doi: 10.2337/db10-1806. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sattar N. Biomarkers for diabetes prediction, pathogenesis or pharmacotherapy guidance? Past, present and future possibilities. Diabet Med. 2012;29(1):5–13. doi: 10.1111/j.1464-5491.2011.03480.x. [DOI] [PubMed] [Google Scholar]
- 11.Krakoff J, Clark JM, Crandall JP, Wilson C, Molitch ME, Brancati FL, et al. Effects of metformin and weight loss on serum alanine aminotransferase activity in the diabetes prevention program. Obesity (Silver Spring) 2010;18(9):1762–1767. doi: 10.1038/oby.2010.21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Rakoski MO, Singal AG, Rogers MA, Conjeevaram H. Meta-analysis: insulin sensitizers for the treatment of non-alcoholic steatohepatitis. Aliment Pharmacol Ther. 2010;32(10):1211–1221. doi: 10.1111/j.1365-2036.2010.04467.x. [DOI] [PubMed] [Google Scholar]
- 13.Retnakaran R, Ye C, Hanley AJ, Harris SB, Zinman B. Discordant effects on central obesity, hepatic insulin resistance, and alanine aminotransferase of low-dose metformin and thiazolidinedione combination therapy in patients with impaired glucose tolerance. Diabetes Obes Metab. 2012;14(1):91–93. doi: 10.1111/j.1463-1326.2011.01481.x. [DOI] [PubMed] [Google Scholar]
- 14.Hermos JA, Cohen SA, Hall R, Gagnon DR, Brophy MT, Fiore LD. Association of elevated alanine aminotransferase with BMI and diabetes in older veteran outpatients. Diabetes Res Clin Pract. 2008;80(1):153–158. doi: 10.1016/j.diabres.2007.11.008. [DOI] [PubMed] [Google Scholar]
- 15.Ruhl CE, Everhart JE. Determinants of the association of overweight with elevated serum alanine aminotransferase activity in the United States. Gastroenterology. 2003;124(1):71–79. doi: 10.1053/gast.2003.50004. [DOI] [PubMed] [Google Scholar]
- 16.National Health and Nutrition Examination Survey (U.S.), Centers for Disease Control and Prevention (U.S.) The National Health and Nutrition Examination Surveys from the National Center for Health Statistics. Atlanta, Ga.: Centers for Disease Control and Prevention; [Google Scholar]
- 17.Association AD. Diagnosis and classification of diabetes mellitus. Diabetes Care. 2012;35(Suppl 1):S64–S71. doi: 10.2337/dc12-s064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.W GE, G LB, M KS. Laboratory procedures used for the Third National Health and Nutrition Examination Survey, 1988–1994. Hyattsville, MD: Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention; 1996. [Google Scholar]
- 19.Ruhl CE, Everhart JE. Upper limits of normal for alanine aminotransferase activity in the United States population. Hepatology. 2012;55(2):447–454. doi: 10.1002/hep.24725. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Zhang X, Geiss LS, Cheng YJ, Beckles GL, Gregg EW, Kahn HS. The missed patient with diabetes: how access to health care affects the detection of diabetes. Diabetes Care. 2008;31(9):1748–1753. doi: 10.2337/dc08-0572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.DeVoe JE, Tillotson CJ, Wallace LS. Usual source of care as a health insurance substitute for U. S. adults with diabetes? Diabetes Care. 2009;32(6):983–989. doi: 10.2337/dc09-0025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Baffy G. Allostasis in Nonalcoholic Fatty Liver Disease: Implications for Risk Assessment. Dig Dis Sci. 2012 doi: 10.1007/s10620-012-2344-8. [DOI] [PubMed] [Google Scholar]
- 23.Wannamethee SG, Shaper AG, Lennon L, Whincup PH. Hepatic enzymes, the metabolic syndrome, and the risk of type 2 diabetes in older men. Diabetes Care. 2005;28(12):2913–2918. doi: 10.2337/diacare.28.12.2913. [DOI] [PubMed] [Google Scholar]
- 24.Mainous AG, Diaz VA, King DE, Everett CJ, Player MS. The relationship of hepatitis antibodies and elevated liver enzymes with impaired fasting glucose and undiagnosed diabetes. J Am Board Fam Med. 2008;21(6):497–503. doi: 10.3122/jabfm.2008.06.080047. [DOI] [PubMed] [Google Scholar]
- 25.Zakim D, Boyer TD. Hepatology : a textbook of liver disease. 4th ed. Philadelphia: Saunders; 2003. [Google Scholar]