Epidemiology/Health Services Research BRIEF REPORT

Low Glycated and Disease in the U.S. Population

1,2 1,2,3 ANDREA L. CHRISTMAN, BA JEANNE M. CLARK, MD, MPH (n = 3,771). To assess the continuous asso- 1,2 1,2 MARIANA LAZO, MD, PHD ELIZABETH SELVIN, PHD, MPH ciations, we fit restricted cubic splines (9). We conducted a sensitivity analysis exclud- ing participants with anemia (hemoglobin d , OBJECTIVE To characterize the association of low HbA1c values ( 4.0%) with liver ,13.5g/dLinmenand,12.0 g/dL in and steatosis. women), presumed iron overload ( d saturation .44% and serum RESEARCH DESIGN AND METHODS Cross-sectional study of 12,533 participants , without aged ,20 years in the Third National Health and Nutrition Examination Survey 10 ng/mL), or B or C (1988–1994). Logistic regression models were adjusted for demographic, lifestyle, and health (n = 11,593, after exclusions). To compare status variables. low HbA1c to low fasting glucose, we mod- eled the 1st percentile versus the rest of the RESULTSdHbA values ranged from 3.2 to 15.7%, and 84 participants had HbA ,4.0% in 1c 1c values for HbA1c (cutpoint, 4.0%) and fast- the population (mean age 44, 52% female, 15% black or Hispanic). We observed J-shaped ing blood glucose (cutpoint, 73.4 mg/dL) in associations between HbA1c and liver enzymes and hepatic steatosis. In adjusted models, HbA1c the morning fasting subsample ( = 8,747). ,4.0% was strongly associated with elevated alanine aminotransferase (OR 3.62 [95% CI 1.09– n 12.02]) and aspartate aminotransferase (6.80 [2.99–15.43]). All analyses accounted for the complex survey design (5). d CONCLUSIONS Low HbA1c values were associated with liver enzymes and steatosis in the U.S. population. may partially explain the association of HbA1c with mortality and other long-term outcomes. RESULTSdCharacteristics of the popu- lation are shown in Supplementary Table 1. Diabetes Care 34:2548–2550, 2011 HbA1c ranged from 3.2 to 15.7%. The prevalence of elevated ALT, AST, or GGT ecent studies show that low HbA1c HbA1c was measured via high- was 5.9, 5.3, and 13.8%, respectively. The Rin nondiabetic individuals is associ- performance liquid chromatography (in- prevalence of hepatic steatosis was 18.8%. ated with increased mortality (1–3). terassay coefficient of variation ,2.0%), We observed J-shaped associations The biologic processes underlying this as- and values ,4.0% or .11.0% were re- between HbA1c andelevatedliveren- sociation remain unclear, although liver peated for verification (7). If hemoglobin zymes and hepatic steatosis (Fig. 1). The disease has been hypothesized (2,4). Our variants were present, affinity chromatog- adjusted odds ratios (ORs) and 95% CIs objective was to examine the association raphy was used (7). We considered ALT, are shown in Supplementary Table 2. In , – between low HbA1c ( 4.0%); elevated AST, and GGT elevated if they were above adjusted models with HbA1c 5.0 5.5% as , liver enzymes alanine aminotransferase the laboratory-defined upper limit of nor- reference group, HbA1c 4.0% was asso- (ALT), aspartate aminotransferase (AST), mal: ALT, .40 units/L for men, .31 ciated with elevated ALT (OR 3.62 [95% and g-glutamyltransferase (GGT); and he- units/L for women; AST, .37 units/L for CI 1.09–12.02]) and AST (6.80 [2.99– , patic steatosis in U.S. adults. men, .31 units/L for women; and GGT, 15.43]). HbA1c 4.0% was also associ- .51 units/L for men, .33 units/L for ated with elevated GGT and with hepatic RESEARCH DESIGN AND women (7). Hepatic steatosis detected by steatosis, but these were not statistically METHODSdThe Third National Health ultrasound was available for all partici- significant. In addition, we found trends and Nutrition Examination Survey (NHANES pants aged 20 to 74 years and was defined for higher total , higher AST-to- III 1988–1994) was a cross-sectional survey as present or absent (8). ALT ratio, lower serum , and lower of the U.S. population (5) and is described in We used logistic regression to examine platelets in individuals with HbA1c ,4.0% detail elsewhere (6,7). We restricted our the association between HbA1c categories (Supplementary Table 1). analyses to 12,533 participants without and elevated liver enzymes and hepatic When we restricted our analysis to self-reported diabetes who had complete steatosis. We repeated our analyses among participants with no history of alcohol , fi information on variables of interest. participants who never consumed alcohol consumption, HbA1c 4.0% was signi - cantly associated with elevated ALT, AST, ccccccccccccccccccccccccccccccccccccccccccccccccc and GGT but not hepatic steatosis. After From the 1Department of Epidemiology, Bloomberg School of Public Health, Johns Hopkins University, excluding participants with anemia, iron Baltimore, Maryland; the 2Welch Center for Prevention, Epidemiology, and Clinical Research, Johns 3 overload, and hepatitis B or C, the ORs Hopkins University, Baltimore, Maryland; and the Department of Medicine, Johns Hopkins University remained elevated but were no longer sta- School of Medicine, Baltimore, Maryland. fi Corresponding author: Andrea L. Christman, [email protected]. tistically signi cant. Received 20 May 2011 and accepted 29 August 2011. HbA1c below the 1st percentile was DOI: 10.2337/dc11-0944 associated with elevated ALT (OR 3.18 This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/suppl/doi:10 [95% CI 1.00–10.27]) and AST (5.80 .2337/dc11-0944/-/DC1. – © 2011 by the American Diabetes Association. Readers may use this article as long as the work is properly [2.55 13.19]) but not with elevated GGT cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.org/ or hepatic steatosis (Supplementary Table licenses/by-nc-nd/3.0/ for details. 3). Fasting glucose below the 1st percentile

2548 DIABETES CARE, VOLUME 34, DECEMBER 2011 care.diabetesjournals.org Christman and Associates

d Figure 1 Adjusted ORs of elevated ALT (A), elevated AST (B), elevated GGT (C), and hepatic steatosis on ultrasound (D)byHbA1c value. The figures show adjusted ORs from cubic spline modes. The shaded area is the 95% CI. The ORs were adjusted for age (years), race/ethnicity (non- Hispanic white; non-Hispanic black; Mexican American), sex (male; female), education (,12 years; 12 years; .12 years), income (below poverty level; above or at poverty level), smoking status (never; former; current #1 pack per day; current .1 pack per day), BMI (kg/m2), history of hypertension, total (mmol/L), HDL cholesterol (mmol/L), C-reactive protein (mg/dL), alcohol consumption (never; former low/mod- erate [,5 drinks/day for men or women]; former high [$5 drinks/day for men or women]; current low [#1 drink/day for women or #2 drinks/day for men]; current moderate/high [.1to,5 drinks per day for women or .2to,5 drinks per day for men]; current high [$5 drinks/day for men or women]), history of , history of nonskin cancer, and health status (excellent; very good; good; fair; poor).

fi was not signi cantly associated with ele- mechanisms by which low HbA1c values demonstrate variability in markers (14) vated ALT, AST, GGT, or hepatic steatosis. may be associated with both abnormal and ultrasound is moderately reliable for liver function and increased mortality in- detecting hepatic steatosis compared with CONCLUSIONSdWe observed J- clude alcohol consumption, abnormal liver biopsy (15). Although the NHANES shaped associations of HbA1c with erythrocyte turnover and function (e.g., III study protocol included repeat HbA1c liver enzymes and hepatic steatosis in anemia and iron overload) (11,12), the testing for extreme HbA1c values and we this sample of U.S. adults (8). Specifi- viral hepatitides (2), and abnormally low performed sensitivity analyses, we were cally, HbA1c values ,4.0% were signif- glycemia (2). We found little evidence not able to assess the potential influence icantly associated with elevated ALT and that this association was mediated by al- of decreased erythrocyte life span per se AST, while associations with hepatic cohol consumption. After excluding par- (), which may alter HbA1c (11). steatosis and elevated GGT were less pro- ticipants with anemia, iron overload, and Strengths of this study include the large nounced. After exclusion of participants hepatitis B or C, the association of HbA1c study population, rigorous data collection with anemia, iron overload, and hepatitis values ,4.0%andelevatedALTandAST procedures, and nationally representative B or C, these associations were no longer were no longer significant, suggesting design. We examined the association of significant. these factors may partially mediate the as- HbA1c with liver disease across the entire New recommendations for the use of sociation.Wedidnotobserveanassoci- spectrum of HbA1c values in nondiabetic HbA1c for diagnosis of diabetes (10) will ation between low fasting glucose and individuals, and we included a compre- likely result in more HbA1c testing. In liver disease, suggesting that the associ- hensive liver disease assessment. light of evidence that very low HbA1c values ationwithlowHbA1c may be independent In conclusion, we observed J-shaped – are associated with total mortality (1 3), it of glycemic pathways. Low HbA1c may be a associations between HbA1c and elevated is important to understand the clinical sig- general marker of poor health, analogous to liver enzymes and hepatic steatosis by ul- nificance of low HbA1c values. Our study low cholesterol levels (13). trasound in a representative sample of the extends prior work in NHANES III, which Certain limitations of this study U.S. population. Additional work is needed reported elevated ALT and AST and a high should be considered when interpreting to determine how liver disease may be re- prevalence of hepatitis C (11.1%) in indi- our results. We had one measurement of lated to the observed associations of low viduals with low HbA1c (2). Hypothesized the liver enzymes, and previous studies HbA1c with total mortality. care.diabetesjournals.org DIABETES CARE, VOLUME 34, DECEMBER 2011 2549 Low HbA1c and liver disease

adults in the U.S. Diabetes Care 2009;32: Centers for Disease Control and Pre- AcknowledgmentsdA.L.C. was supported by 1440–1446 vention. Available from http://www.cdc. National Institutes of Health (NIH)/National 4. McNeely M. Case study: diabetes in a pa- gov/nchs/data/nhanes/nhanes3/Hepatic_ Institute of Diabetes and Digestive and Kidney tient with . Clin Diabetes 2004; Steatosis_Ultrasound_Procedures_Manual. Diseases (NIDDK) Training Grant T32-DK- 22:42–43 pdf. Accessed 7 January 2011 062707. M.L. was supported by NIH/NIDDK 5. Analytical and reporting guidelines: The 9. Harrell FE. Regression Modeling Strategies Grant R01-DK-083393 and by the American Third National Health and Nutrition Ex- With Applications to Linear Models, Logistic Diabetes Association. J.M.C. was supported by amination Survey, NHANES III (1988– Regression, and Survival Analysis. New York, NIH/NIDDK Grant R01-DK-083393. E.S. was 1994) [article online], 1996. Hyattsville, Springer, 2001 supported by NIH/NIDDK grants K01-DK- MD, National Center for Health Statistics, 10. American Diabetes Association. Diagnosis 076595 and R01-DK-089174. Centers for Disease Control and Pre- and classification of diabetes mellitus. Di- No potential conflicts of interest relevant to vention. Available from http://www.cdc. abetes Care 2010;33(Suppl. 1):S62–S69 this article were reported. gov/nchs/data/nhanes/nhanes3/nh3gui. 11. Cohen RM, Franco RS, Khera PK, et al. A.L.C. analyzed data, contributed to dis- pdf. Accessed 7 January 2011 Red cell life span heterogeneity in hema- cussion, wrote the manuscript, and reviewed 6. National Center for Health Statistics. Plan tologically normal people is sufficient and edited the manuscript. M.L. and J.M.C. and Operation of the Third National Health to alter HbA1c. Blood 2008;112:4284– researched data, contributed to discussion, and Nutrition Examination Survey, 1988– 4291 and reviewed and edited the manuscript. E.S. 1994. Washington, DC, U.S. Govt. Print- 12. Ford ESCC, Cowie CC, Li C, Handelsman Y, contributed to discussion and reviewed and ing Office, 1994 Bloomgarden ZT. Iron-deficiency anemia, edited the manuscript. 7. Laboratory procedures used for the Third non-iron-deficiency anemia and HbA1c National Health and Nutrition Examination among adults in the US. J Diabetes 2011; Survey (NHANES III), 1988–1994 [article 3:67–73 online], 1996. Bethesda, MD, U.S. De- 13. Volpato S, Zuliani G, Guralnik JM, Palmieri E, References partment of Health and Human Services, Fellin R. The inverse association between age 1. Selvin E, Steffes MW, Zhu H, et al. Gly- Centers for Disease Control and Prevention, and cholesterol level among older patients: cated hemoglobin, diabetes, and cardio- National Center for Environmental Health, the role of poor health status. Gerontology vascular risk in nondiabetic adults. N Engl National Center for Health Statistics. Avail- 2001;47:36–45 J Med 2010;362:800–811 able from http://www.cdc.gov/nchs/data/ 14. Lazo M, Selvin E, Clark JM. Brief commu- 2. Carson AP, Fox CS, McGuire DK, et al. Low nhanes/nhanes3/cdrom/nchs/manuals/ nication: clinical implications of short-term hemoglobin A1c and risk of all-cause mor- labman.pdf. Accessed 7 January 2011 variability in liver function test results. Ann tality among US adults without diabetes. 8. Third National Health and Nutrition Ex- Intern Med 2008;148:348–352 Circ Cardiovasc Qual Outcomes 2010;3: amination Survey: hepatic steatosis ultra- 15. Gomercic M, Duvnjak M, Barsic N. Ultra- 661–667 sound images assessment procedures sonography in the diagnosis of nonalcoholic 3. Saydah S, Tao M, Imperatore G, Gregg E. manual [article online], 2011. Hyattsville, fatty liver disease. Acta Med Croatica 2009; GHb level and subsequent mortality among MD, National Center For Health Statistics, 63(Suppl. 3):1–3

2550 DIABETES CARE, VOLUME 34, DECEMBER 2011 care.diabetesjournals.org