1878 Diabetes Care Volume 37, July 2014

Sneha B. Sridhar, Fei Xu, Jeanne Darbinian, Pregravid Liver Enzyme Levels Charles P. Quesenberry, Assiamira Ferrara, and Risk of Gestational Diabetes and Monique M. Hedderson Mellitus During a Subsequent Pregnancy Diabetes Care 2014;37:1878–1884 | DOI: 10.2337/dc13-2229

OBJECTIVE Liver enzymes are independent predictors of type 2 diabetes. Although liver fat content correlates with features of insulin resistance, a risk factor for developing gestational diabetes mellitus (GDM), the relationship between liver enzymes and GDM is unclear. The objective of this study was to assess whether pregravid liver enzyme levels are associated with subsequent risk of GDM.

EPIDEMIOLOGY/HEALTH SERVICES RESEARCH RESEARCH DESIGN AND METHODS A nested case-control study was conducted among women who participated in the Kaiser Permanente Northern California multiphasic health checkup (1984– 1996) and had a subsequent pregnancy (1984–2009). Case patients were 256 women who developed GDM. Two control subjects were selected for each case patient and matched for year of blood draw, age at examination, age at preg- nancy, and number of intervening pregnancies.

RESULTS Being in the highest quartile versus the lowest quartile of g-glutamyl transferase (GGT) levels was associated with a twofold increased risk of subsequent GDM (odds ratio 1.97 [95% CI 1.14–3.42]), after adjusting for race/ethnicity, prepregnancy BMI, family history of diabetes, and alcohol use. This result was attenuated after adjusting for homeostasis model assessment of insulin resistance (HOMA-IR), fasting status, and rate of gestational weight gain. There was significant interaction between GGT and HOMA-IR; the association with GGT was found among women in the highest tertile of HOMA-IR. Aspartate aminotransferase and alanine aminotransferase were not associated with increased GDM risk.

CONCLUSIONS Pregravid GGT level, but not alanine aminotransferase or aspartate aminotransfer- ase level, predicted the subsequent risk of GDM. Markers of liver fat accumulation, such as GGT level, are present years before pregnancy and may help to identify Division of Research, Kaiser Permanente Northern women at increased risk for subsequent GDM. California, Oakland, CA Corresponding author: Sneha B. Sridhar, sneha.x. [email protected]. Gestational diabetes mellitus (GDM), defined as carbohydrate intolerance with first – – Received 20 September 2013 and accepted 9 onset or recognition in pregnancy, affects 4 7% of women in the U.S. (1 3). GDM is March 2014. associated with adverse perinatal outcomes and is a risk factor for the development © 2014 by the American Diabetes Association. of type 2 diabetes in both the mother and her offspring. Strategies to prevent GDM See http://creativecommons.org/licenses/by- hold great potential as a means by which to prevent or delay the onset of diabetes. nc-nd/3.0/ for details. care.diabetesjournals.org Sridhar and Associates 1879

Insulin resistance and inadequate insu- KPNC subscribers are representative of Registry captures the results of all lin response are two known mechanisms the region (12). screening and diagnostic tests for GDM underlying the pathophysiology of both The source population consisted of from the KPNC electronic laboratory da- GDM and type 2 diabetes. female KPNC members who completed a tabase (data available since 1994). Laboratory tests for g-glutamyl trans- voluntary multiphasic health checkup Study Design ferase (GGT), alanine aminotransferase (MHC) at the Kaiser Permanente Oakland This is a nested case-control study, within a (ALT), and aspartate aminotransferase Medical Center between 1984 and 1996. cohort of 4,098 women who took part in (AST) are commonly performed to as- KPNC members at this facility were invited an MHC examination, had an extra tube of sess the overall health of the liver. Liver to complete a comprehensive health serum stored for future use, and had a sub- fat content has been shown to correlate checkup upon study enrollment. The sequent pregnancy, on average, 7 years with features of insulin resistance inde- MHC consisted of a clinic visit for the com- after the MHC examination. All cohort pendent of BMI and abdominal obesity. pletion of questionnaires and clinical members in whom GDM subsequently de- While the gold standard for measuring measurements, including blood pressure, veloped were included as case patients; nonalcoholic fatty liver disease is a weight, and serum glucose and choles- two control subjects were selected for liver biopsy, testing of ALT, AST, and terol levels (measured in serum obtained each case patient from among women GGT seems to provide reasonable non- from a random blood draw). An extra not meeting the GDM case definition. invasive surrogate measures for use in serum sample was collected and stored fi epidemiologic studies (4). The liver is at 2408C for future use. The goal of the GDM Case De nition We identified 267 women with GDM ac- crucial to maintaining glucose homeo- MHC was to provide health maintenance cording to the KPNC electronic data- stasis, both during fasting and postpran- through early diagnosis (13). BMI was cal- bases. Case patients had either 1) dial states, and thereby plays a role in the culated as kilograms per square meter; glucose values obtained during a stan- development of type 2 diabetes. Serum height and weight were measured dard 100-g, 3-h oral glucose tolerance GGT level is also a marker of oxidative using a stadiometer and a balance beam test (OGTT) that met the Carpenter- stress (5,6). Oxidative stress is the condi- scale, respectively. Information on age, Coustan plasma glucose thresholds for tion of increased free radical activity and sex, race/ethnicity, education level, ciga- GDM (as outlined by the American Col- high lipid oxidation, and it plays a role in rette smoking, family history of diabetes, lege of Obstetricians and Gynecologists) the etiology of type 2 diabetes by induc- medical history, alcohol consumption (15) in the laboratory database (n = ing insulin resistance in the peripheral ($1vs.,1 drink/day), coffee consump- 228), or 2) a hospital discharge diagnosis tissues and impairing insulin secretion tion, use of medications, and hours since of GDM in the electronic hospital dis- from the pancreatic b-cells (7,8). GGT last food ingestion was collected using charge database for pregnancies occur- catabolizes extracellular glutathione self-administered questionnaires (13). Se- ring before the electronic laboratory (GSH), which has an antioxidant func- rum glucose was measured in serum ob- data were available (prior to 1994; n = tion; therefore, GGT levels may become tained from a random blood draw using 39). Standardized medical chart review elevated in order to produce more GSH the hexokinase method, and total choles- was conducted by trained abstractors to in response to oxidative stress (9). terol was assessed by the regional labo- confirm that these 267 women had 100-g, Past research suggests that during ratory of KPNC at the time of the MHC 3-h OGTT results meeting the Carpenter- a normal pregnancy, liver enzyme levels using a Kodak Ektachem chemistry ana- Coustan criteria (15) for GDM (plasma may change in response to the increased lyzer. This laboratory participates in the glucose thresholds: fasting 5.3 mmol/L insulin resistance induced by pregnancy College of American Pathologists accred- [95 mg/dL]; 1-h 10.0 mmol/L [180 (10,11); therefore, it is important to de- itation and monitoring program. mg/dL]; 2-h 8.6 mmol/L [155 mg/dL]; termine whether prepregnancy levels of Among women 15–45 years of age 3-h 7.8 mmol/L [140 mg/dL]). Case liver enzymes are related to subsequent (median age 34 years) who participated patients were excluded if, at the time risk of GDM, in order to clarify the tempo- in the MHC from 1985 to 1996 (n = of the MHC examination, they had a ran- ral sequence of the association. Although 27,743 with clinical and questionnaire dom glucose levels of .200 mg/dL (n = liver enzymes are known to correlate with data, as well as an extra serum sample), 6), no indication of GDM during the in- features of insulin resistance, a risk factor we identified 4,098 women who sub- dex pregnancy (n = 4), or they had impaired for the development of GDM, the relation- sequently delivered an infant by 2010 glucose tolerance with insufficient follow- ship between prepregnancy liver enzyme by searching the KPNC hospitalization up testing (n = 1), leaving a total of 256 levels and GDM is unclear. Thus, the aim database and the Pregnancy Glucose confirmed cases of GDM. of this case-control study is to examine the Tolerance and GDM Registry (3), an ac- association between prepregnancy ALT, tive surveillance registry that annually Control Selection and Matching AST, and GGT levels and the risk of the identifies all pregnancies resulting in a Criteria subsequent development of GDM. livebirth or stillbirth among KPNC mem- Among women without an indication of bers. Women with recognized pregravid GDM, control subjects were randomly RESEARCH DESIGN AND METHODS diabetes (14) are excluded from the selected; two control subjects were in- The setting is Kaiser Permanente Northern GDM Registry; therefore, women in dividually matched to each case patient California (KPNC), an integrated health whom diabetes had been diagnosed based on the year of MHC serum collec- care delivery system that provides medical prior to the index pregnancy were not tion date (63 months), age at MHC se- care for about one-third of the underlying eligible to be included in the study. The rum collection (62 years), the number population in the San Francisco Bay Area. Pregnancy Glucose Tolerance and GDM of intervening pregnancies (0, 1, $2), 1880 Pregravid Liver Enzymes and GDM Diabetes Care Volume 37, July 2014

and age at delivery of the index preg- the relative risk of GDM in relation to respectively; P value ,0.001). The nancy (62 years). We matched for the prepregnancy GGT, ALT, and AST levels. mean prepregnancy AST level was also year of serum collection to account for Women were categorized by quartile of higher among women in whom GDM de- any potential degradation in the quality GGT,ALT,andASTlevels,asdefined veloped versus those in whom it did not of the serum over time, thereby assuring among control subjects. Variables eval- (13.9 vs. 11.8 units/L), although the that the sample storage time was ap- uated for confounding included race/ difference was not significant. proximately the same for case patients ethnicity, pregravid BMI (in kilograms Table 2 shows the ORs and 95% CIs for and control subjects. Since GDM is more per square meter), parity, cigarette risk of GDM determined by pregravid liver common in older women, we matched smoking, alcohol use, maternal educa- enzyme levels. The first model adjusted on age at serum collection and age at tion, and family history of diabetes, all for race/ethnicity, BMI, family history of delivery. We matched on the number assessed at the time of liver enzyme diabetes, and alcohol use at the time of pregnancies to account for any differ- measurement. To examine the effect of the MHC. For GGT, there was a trend ences in pregnancies between the initial of weight gain during pregnancy up to of increasing risk of GDM as the quartile examination and the index pregnancy. the time of GDM diagnosis, we added increased. Being in the fourth versus the Control subjects were excluded from the this variable to the fully adjusted condi- first quartile of GGT resulted in a twofold analysis if they had glucose values that tional logistic regression model. increase in the odds of the development were diagnostic of GDM found during To assess confounding, we entered of GDM, and the association was signifi- medical chart abstraction (n = 5); an ab- covariates into a logistic regression cant (OR 1.97 [95% CI 1.14–3.42]) (Table normal screening glucose level but no fol- model, one at a time, and compared 2). After further adjusting for HOMA-IR low-up diagnostic glucose test (n = 5); or the adjusted and unadjusted estimates. (in tertiles), fasting status ($6hsince one abnormal glucose value on the diag- We included covariates that altered un- the last food at the time of MHC exami- nostic glucose test (n = 5), suggestive of adjusted estimates by $10%. nation), and rate of gestational weight “mild” GDM. Of the 512 matched control To assess the potential modifying gain (in tertiles) among the full cohort, subjects identified, 497 were eligible. effects of the prepregnancy tertile of the association was no longer significant HOMA-IR (dichotomized as $67th per- (OR 1.57 [95% CI 0.84–2.93]). Similar re- Exposure Variables centile vs. ,67th percentile), prepreg- sultswerefoundwhenrestrictedtowomen Serum Biomarker Assays nancy BMI, racial/ethnic group, and who were fasting for $6h.Wealso Serum samples were thawed, aliquoted, time since MHC examination, we in- conducted a sensitivity analysis excluding and transported in batches on dry ice to cluded appropriate cross-product (inter- women who drank one or more alcoholic the laboratory of Dr. Peter Havel at the action) terms in regression models. This drinks per day at the time of the MHC ex- University of California, Davis, for anal- study was approved by the human amination, and the results were similar. ysis. GGT, ALT, and AST were measured subjects committee of the Kaiser Foun- HOMA-IR was calculated both in on a Poly-Chem analyzer (MedTest DX, dation Research Institute. women who had fasted for at least 6 h Cortlandt Manor, NY). The intra-assay (n = 419) and in those who had not and interassay coefficients of variation RESULTS fasted (n = 306). There were no signifi- were 4.7% and 5.6% (GGT), 4.5% and Table 1 summarizes the demographic, cant differences by fasting status in cor- 11.7% (ALT), and 2.7% and 9.5% (AST), anthropometric, reproductive, and met- relations between HOMA-IR and BMI, respectively. Insulin was measured abolic characteristics of the study partic- GGT, ALT, AST, and glucose levels. with a radioimmunassay (Millipore). ipants, by case/control status. Women Among women who were fasting, the The intra-assay and interassay coeffi- in whom GDM developed were more correlations between HOMA-IR and cients of variation were ,4.0% and likely to have ,12 years of education, the liver enzymes were as follows: GGT ,10%, respectively. Insulin resistance to be Asian or Hispanic, to have two or r = 0.32, P , 0.0001; AST r =0.10,P = was calculated based on the homeosta- more children at the time of the MHC, to 0.05; and ALT r =0.22,P , 0.0001. sis model assessment of insulin resis- abstain from alcohol, and to have a fam- Among women who were not fasting, tance (HOMA-IR) using the following ily history of type 2 diabetes compared the correlations between HOMA-IR equation: (fasting glucose 3 fasting in- with women in whom GDM did not and the liver enzymes were as follows: sulin)/22.5, where glucose was mea- develop. Women in whom GDM devel- GGT r =0.16,P , 0.01; AST r =0.09,P = sured in millimoles per liter and insulin oped also had higher levels of several 0.12; and ALT r =0.18,P , 0.01. While in milliunits per milliliter (16). cardiometabolic risk factors at the women who were not fasting for $6h MHC examination, including BMI, serum had higher HOMA-IR levels, there Rate of Gestational Weight Gain per Week glucose, total cholesterol, systolic and remained a significant difference be- The rate of gestational weight gain per diastolic blood pressure, and serum in- tween case patients and control sub- week was calculated as follows: (weight sulin concentrations, and weight gain jects ([mean 6 SD] women fasting ,6h: measured at or before the glucose from the MHC examination to the index case patients 9.0 6 11.5; control sub- screening test 2 prepregnancy weight)/ pregnancy. Mean prepregnancy GGT jects 4.5 6 3.8; P , 0.001; women fast- weeks of gestation attained at the time of and ALT levels (in units per liter) were ing $6h:casepatients4.16 3.5; the weight measurement. significantly higher in women in whom control subjects 2.9 6 2.9; P , 0.001); Statistical Analysis GDM developed, when compared with therefore, we chose to include everyone Conditional logistic regression was used those in whom GDM did not develop in the analytic cohort regardless of to obtain odds ratios (ORs) to estimate (28.0 vs. 22.4 and 8.5 vs. 6.7 units/L, fasting status. care.diabetesjournals.org Sridhar and Associates 1881

Table 1—Characteristics of case patients and control subjects Neither ALT nor AST was associated Control with an increased risk of GDM, and no GDM case patients subjects clear trend was observed. The interac- Characteristics (n =256) (n =497) P value tion between BMI and race/ethnicity did Age at MHC examination 28.2 6 5.5 28.4 6 5.2 0.78a not reach statistical significance. Age at delivery (years) 35.4 6 5.1 35.1 6 4.9 0.43b Figure 1 displays the ORs for GDM ,30 39 (15.2) 80 (16.1) stratified by prepregnancy tertile of 30–34 73 (28.5) 145 (29.2) $ 35–39 102 (39.8) 183 (36.8) HOMA-IR, dichotomized ( 67th per- $40 42 (16.4) 89 (17.9) centile vs. ,67th percentile). There Time between examination and delivery 7.1 6 4.4 6.7 6 4.4 0.21a was significant interaction by prepreg- Education (years) 0.24b nancy HOMA-IR (P = 0.082), and the as- #12 74 (28.9) 119 (23.9) sociations with being in the top three 13–15 85 (33.2) 157 (31.6) $16 92 (35.9) 214 (43.1) quartiles of GGT level were stronger Unknown 5 (2.0) 7 (1.4) for women in the highest tertile of Race/ethnicity ,0.001b HOMA-IR ($67th percentile) before Non-Hispanic White 50 (19.5) 186 (37.4) pregnancy (Quartile 2 OR 2.71 [95% CI African American 91 (35.5) 184 (37.0) 1.05–7.01]; Quartile 3 3.78 [1.42– Asian/Pacific Islander 80 (31.3) 84 (16.9) Hispanic 35 (13.7) 43 (8.7) 10.08]; and Quartile 4 4.93 [1.93– Parity ,0.001b 12.60] vs. Quartile 1 1.00). No significant 0 142 (55.5) 278 (55.9) associations were observed among 1 47 (18.4) 106 (21.3) women in the lower two tertiles of $2 44(17.2) 70(14.1) Unknown 23 (9.0) 43 (8.7) HOMA-IR. Gestational age at birth (weeks) 0.01b Figure 2 displays the ORs for GDM strat- $37 218 (84.8) 460 (90.7) ified by prepregnancy BMI, dichotomized ,37 39 (15.2) 39 (7.7) (BMI $25 vs. ,25 kg/m2). There was Large-for-gestational age at birthc ,0.01b increasing risk of GDM as the tertile of No 198 (81.1) 427 (89.5) Yes 46(18.9) 50(10.5) GGT increased, and being in the highest Alcohol use ,0.001b tertile of GGT with a prepregnancy BMI 2 None 74 (28.9) 81 (16.3) of ,25 kg/m resulted in the greatest Occasional or more drinks/day 149 (58.2) 346 (69.6) risk of GDM (OR 2.44 [95% CI 1.23– Unknown 33 (12.9) 70 (14.1) 4.83]). Smoking 0.40b Never 150 (58.6) 277 (55.7) CONCLUSIONS Former 37 (14.5) 92 (18.5) Current 38 (14.8) 61 (12.3) In this case-control study, pregravid GGT, Unknown 31 (12.1) 67 (13.5) but not ALT or AST, was found to be as- b Hypertension status at index pregnancy ,0.001 sociated with an increased risk of the de- No hypertension 138 (53.9) 326 (65.5) Preexisting hypertensiond 28 (10.9) 18 (3.6) velopment of GDM. The association Gestational hypertension 33 (12.9) 68 (13.7) appeared to be moderated by increased Preeclampsia 42 (16.4) 37 (7.4) insulin resistance, and in the stratified Family history of diabetes 151 (59.0) 192 (38.6) ,0.001b analysis, it was present only among 2 a BMI (kg/m ) 26.0 6 6.5 23.7 6 4.6 ,0.001 women who were in the top tertile of a Weight change from MHC to pregnancy (kg) 8.9 6 9.9 4.4 6 8.2 ,0.001 HOMA-IR before pregnancy. e 6 6 , a Rate of gestational weight gain (kg/week) 0.3 0.2 0.4 0.2 0.07 Our findings with GGT are consistent 6 6 , a Serum glucose (mg/dL) 89.6 13.5 83.6 8.3 0.001 with previous research examining liver 6 6 , a Serum cholesterol (mg/dL) 182.9 33.2 176 32.6 0.01 enzymes and type 2 diabetes. Several Systolic blood pressure (mmHg) 115.6 6 14.7 113.3 6 13.4 ,0.05a cohort studies have found that higher Diastolic blood pressure (mmHg) 69.9 6 10.4 68.3 6 9.0 ,0.05a GGT levels predict the development of White blood cell count (1,000 cells per mm3)6.96 1.9 6.5 6 1.9 ,0.01a type 2 diabetes (17–22). The association GGT (units/L) 28.0 6 21.7 22.4 6 16.6 ,0.001f ALT (units/L) 8.5 6 9.5 6.7 6 3.8 ,0.001f between GGT and type 2 diabetes has AST (units/L) 13.9 6 25.3 11.8 6 6.6 0.18f been found to have a dose-response re- HOMA-IR index 4.1 6 3.5 2.9 6 2.9 ,0.001a lationship (19,21) and to be indepen- Insulin (mU/mL) 25.8 6 28.6 17.5 6 16.7 ,0.001f dent of known risk factors for diabetes (17,20). One study (23) found that GGT Data are mean 6 SD or N (%), unless otherwise indicated. at Test to compare differences in mean values of continuous variables except as noted below for Wilcoxon test. bx2 Test for categorical concentrations were independently as- variables. cSubset of women with singleton births; large-for-gestational age .90th percentile sociated with the risk of prediabetes and based on race and gestational age-specific quantiles. dIncludes women who experienced were positively associated with insulin e preeclampsia superimposed on preexisting hypertension. Weight change in kilograms per week resistance. Few studies have examined from beginning of index pregnancy until screening glucose (measurement obtained 1 h after the 50-g oral challenge). Data were available for 235 case patients and 446 control subjects. liver enzyme levels during pregnancy in fWilcoxon test for differences in median values. relation to GDM risk, and findings have been inconsistent. In a study of 2,610 1882 Pregravid Liver Enzymes and GDM Diabetes Care Volume 37, July 2014

Table 2—Association of GDM with prepregnancy ALT, AST, and GGT levels, from conditional logistic regression models Conditional logistic regression models Case patients Control subjects Prepregnancy risk factor (n =256) (n =497) Crude Multivariable-adjusted1 Multivariable-adjusted2 ALT (units/L) Quartile 1 (0.5–3.9) 44 (17.2) 124 (25.0) 1.00 1.00 1.00 Quartile 2 (4.0–5.9) 61 (23.8) 124 (25.0) 0.90 (0.55–1.47) 0.79 (0.45–1.39) 0.80 (0.43–1.50) Quartile 3 (6.0–8.4) 70 (27.3) 124 (25.0) 1.25 (0.76–2.04) 1.04 (0.59–1.82) 1.18 (0.62–2.23) Quartile 4 (8.5–35.0) 81 (31.6) 125 (25.0) 1.55 (0.92–2.59) 1.10 (0.61–1.99) 1.12 (0.57–2.23) AST (units/L) Quartile 1 (2.0–7.9) 73 (28.5) 124 (25.0) 1.00 1.00 1.00 Quartile 2 (8.0–10.4) 53 (20.7) 124 (25.0) 0.63 (0.40–0.97) 0.47 (0.28–0.79) 0.38 (0.21–0.68) Quartile 3 (10.5–14.4) 58 (22.7) 124 (25.0) 0.73 (0.45–1.18) 0.62 (0.36–1.07) 0.51 (0.27–0.96) Quartile 4 (14.5–92.0) 72 (28.1) 125 (25.0) 0.84 (0.48–1.45) 0.59 (0.31–1.11) 0.53 (0.25–1.13) GGT (units/L) Quartile 1 (8.0–13.4) 37 (14.5) 124 (25.0) 1.00 1.00 1.00 Quartile 2 (13.5–17.9) 47 (18.4) 124 (25.0) 1.41 (0.84–2.36) 1.28 (0.73–2.25) 1.20 (0.63–2.26) Quartile 3 (18.0–24.9) 74 (28.9) 124 (25.0) 2.03 (1.23–3.37) 1.72 (0.99–2.97) 1.55 (0.84–2.87) Quartile 4 (25.0–173.0) 98 (38.3) 125 (25.0) 2.91 (1.77–4.77) 1.97 (1.14–3.42) 1.57 (0.84–2.93) Data are n (%) or OR (95% CI). 1Adjusted for race/ethnicity, prepregnancy BMI, family history of diabetes, and alcohol use at time of the MHC examination (one or more vs. less than one drink/day). 2Further adjusted for HOMA-IR (in tertiles), fasting status (defined as $6 h since ingestion of the last food at the time of the MHC examination), and rate of gestational weight gain up to screening test (in tertiles).

women in Malaysia in 2012 (24), GGT, the liver, independent of peripheral in- with obesity to impact diabetes risk. ALT, and AST levels were measured at sulin sensitivity or resistance (23). A GGT activity could reflect the formation the time of the 50-g glucose challenge 2011 study by Bonnet et al. (27) found of GSH conjugates during xenobiotic me- test, prior to the 2-h, 75-g OGTT. The that increased plasma GGT level, even tabolism. The authors recommend pro- study found no association between within the normal reference range, spective studies and toxicological the levels of pregnancy liver enzymes was a biomarker of both systemic and studies to test their hypotheses. and the risk of GDM. Another study hepatic insulin resistance, as well as in- In the current study, neither ALT nor (25) assessed GGT levels at the time of creased insulin secretion and decreased AST was associated with GDM. AST is the OGTT and found that they were pos- hepatic insulin clearance among both present throughout the body and levels itively correlated with the 2-h glucose healthy men and women in a nonpreg- can be elevated in a multitude of clinical level. Additionally, increased GGT level nant, nondiabetic state. The authors disorders (30); it has also not been con- was significantly associated with GDM concluded that even a minimal increase sistently associated with diabetes (24). risk in the multivariable logistic regres- in GGT level could serve as an indirect ALT is found primarily in the liver and is sion analysis. The authors of the two marker of enhanced hepatic insulin re- thought to be a marker of liver fat accu- aforementioned studies justified the sistance and impaired glucose disposal mulation (31). The lack of association use of liver enzymes levels during preg- in skeletal muscles (27). Our finding that between ALT level and GDM in the cur- nancy (on average, up to 2 weeks before elevated pregravid GGT levels resulted rent study is surprising, given that ALT the OGTT) for the purpose of predicting in an increased risk of GDM, to a greater has been considered a marker of risk for the development of GDM, given that el- degree among women in the highest the development of type 2 diabetes evated transaminase levels are a chronic tertile of HOMA-IR, lends support to (24); however, several studies have reflection of diabetes risk, “predictive of the hypothesis that insulin resistance found no association between ALT and events sometimes years ahead” (24). To underlies this association. Additionally, type 2 diabetes (32,33). Additionally, a our knowledge, no study other than the elevated GGT level could be a response 2013 meta-analysis (34) concluded that current study has examined the associ- to oxidative stress, which plays a critical associations of liver aminotransferase ation between prepregnancy liver en- role in the pathogenesis of diabetes by levels and type 2 diabetes risk appear zyme levels and GDM. Liver enzyme impairing insulin secretion. Elevations in to have been overestimated in previous levels are relatively inexpensive to mea- GGT, thought to be produced in part as a studies. It is possible that the lack of sure and can be used to identify women result of oxidative stress, result in in- association is due to the relatively who are at risk for the development of creased transport of the tripeptide younger age of our study participants, GDM. GSH into cells, where it can protect the as older age may be a risk factor for Among patients without hepatitis, in- cells from oxidative damage (28). It is increased susceptibility to certain liver creased levels of GGT indicate deposi- thought that GGT level may also be a diseases. Alternatively, the lack of asso- tion of excess fat in the liver (26), marker of exposure to certain environ- ciation may be due to a different mech- which is known to be characterized by mental pollutants. Lee et al. (29) hypoth- anistic effect of ALT in the development insulin resistance (23). Some research- esize that persistent organic pollutants of GDM and type 2 diabetes. Further ers speculate that elevations in liver en- may reside in adipose tissue and act as studies are needed. zyme levels may reflect an underlying endocrine disruptors, and the persistent The strengths of this study include insulin resistance that is localized to organic pollutants may further interact our ability to exclude women with care.diabetesjournals.org Sridhar and Associates 1883

Figure 1—ORs for association between pregravid liver enzyme levels and HOMA-IR and the risk of GDM. glucose values indicative of recognized unclear which one is the antecedent, within 10 years of follow-up (35). This pregestational diabetes. We had the and this represents an avenue on which supports the hypothesis that excess unique ability to look at liver enzyme future studies should focus. The study liver fat is one potential biologic mech- levels measured, on average, 7 years was limited by the lack of data on anism explaining the association be- before pregnancy in a large number of hepatic insulin resistance, liver fat con- tweenGGTandGDM. GDM case patients and matched control tent, and body fat mass and distribu- Based on our findings, the liver en- subjects in a diverse cohort. Assessing tion; the latter might have provided zyme GGT, which is known to be associ- the liver enzyme concentrations before insight as to whether the association ated with an increased risk of the pregnancy is vital to establishing the between GGT and GDM was mediated development of type 2 diabetes, ap- temporal sequence of the association by visceral fat. Past research indicates pears to also be associated with subse- between liver enzymes and GDM, that fatty liver is associated with insulin quent GDM, which is characterized by particularly given the fact that liver en- resistance and inflammation in women decreased insulin sensitivity and in- zyme concentrations may change dur- with a history of GDM, and higher levels creased insulin resistance. This study, ing pregnancy, in part due to increases of biomarkers (as measured by a vali- which is, to our knowledge, the first of in sex steroid levels, which affect metabolic dated fatty liver index) estimating ex- its kind, demonstrates that elevations in and hepatic functions (10). While strong cess liver fat in these women was GGT levels may be present many years evidence links elevated liver enzyme associated with an increased risk of before a pregnancy characterized by concentrations to insulin resistance, it is the development of type 2 diabetes GDM. Monitoring GGT levels before

Figure 2—ORs for association between pregravid liver enzyme levels and BMI and the risk of GDM. 1884 Pregravid Liver Enzymes and GDM Diabetes Care Volume 37, July 2014

pregnancy may help to identify women gene tell us about its functions? Am J Pathol Artery Risk Development in Young Adults (CARDIA) whoareatincreasedriskforthesub- 1995;147:1175–1185 Study. Clin Chem 2003;49:1358–1366 sequent development of GDM. 7. Oberley LW. Free radicals and diabetes. Free 22. Nakanishi N, Suzuki K, Tatara K. Serum Radic Biol Med 1988;5:113–124 gamma-glutamyltransferase and risk of meta- 8. Ceriello A, Motz E. Is oxidative stress the bolic syndrome and type 2 diabetes in middle- pathogenic mechanism underlying insulin resis- aged Japanese men. Diabetes Care 2004;27: Acknowledgments. The authors thank Monica tance, diabetes, and cardiovascular disease? 1427–1432 Highbaugh, Mamie Ford, and Jean Lee, all from The common soil hypothesis revisited. Arterio- 23. Gao F, Pan JM, Hou XH, et al. Liver enzymes Kaiser Permanente Northern California, Division scler Thromb Vasc Biol 2004;24:816–823 concentrations are closely related to prediabe- of Research, for their help in reviewing medical 9. Zhang H, Forman HJ, Choi J. Gamma-glutamyl tes: findings of the Shanghai Diabetes Study II records. transpeptidase in glutathione biosynthesis. (SHDS II). Biomed Environ Sci 2012;25:30–37 Funding. A.F. was supported by National Methods Enzymol 2005;401:468–483 24. Tan PC, Aziz AZ, Ismail IS, Omar SZ. Gamma- Institutes of Health grant P30-DK-092924. This 10. Bacq Y, Zarka O, Brechot´ JF, et al. Liver func- glutamyltransferase, alanine transaminase and research was also supported by Eunice Kennedy tion tests in normal pregnancy: a prospective aspartate transaminase levels and the diagnosis Shriver National Institute of Child Health and study of 103 pregnant women and 103 matched of gestational diabetes mellitus. Clin Biochem Human Development grant R01-HD-065904 to controls. Hepatology 1996;23:1030–1034 2012;45:1192–1196 M.M.H. 11. Moniz CF, Nicolaides KH, Bamforth FJ, 25. Tan PC, Mubarak S, Omar SZ. Gamma- Duality of Interest. No potential conflicts of Rodeck CH. Normal reference ranges for bio- glutamyltransferase level in pregnancy is an in- interest relevant to this article were reported. chemical substances relating to renal, hepatic, dependent risk factor for gestational diabetes Author Contributions. S.B.S. assisted with and bone function in fetal and maternal plasma mellitus. J Obstet Gynaecol Res 2008;34:512–517 data collection and wrote and edited the throughout pregnancy. J Clin Pathol 1985;38: 26. Inoue K, Matsumoto M, Miyoshi Y, manuscript. F.X. and J.D. conducted the data 468–472 Kobayashi Y. Elevated liver enzymes in women analysis and contributed to the writing of the 12. Go AS, Hylek EM, Phillips KA, et al. Preva- with a family history of diabetes. Diabetes Res manuscript. C.P.Q. contributed to the study lence of diagnosed atrial fibrillation in adults: Clin Pract 2008;79:e4–e7 design, provided statistical expertise, and con- national implications for rhythm management 27. Bonnet F, Ducluzeau PH, Gastaldelli A, et al.; tributed to the writing of the manuscript. A.F. and stroke prevention: the AnTicoagulation and RISC Study Group. Liver enzymes are associated designed the study and edited the manuscript. Risk Factors in Atrial Fibrillation (ATRIA) Study. with hepatic insulin resistance, insulin secretion, M.M.H. designed the study, oversaw the data JAMA 2001;285:2370–2375 and glucagon concentration in healthy men and collection, and wrote and edited the manu- 13. Collen MF. Multiphasic Health Testing Ser- women. Diabetes 2011;60:1660–1667 script. M.M.H. is the guarantor of this work and, vices. New York, John Wiley & Sons, 1978 28. McLennan SV, Heffernan S, Wright L, et al. as such, had full access to all the data in the 14. Selby JV, Ray GT, Zhang D, Colby CJ. Excess Changes in hepatic glutathione metabolism in study and takes responsibility for the integrity costs of medical care for patients with diabetes diabetes. Diabetes 1991;40:344–348 of the data and the accuracy of the data in a managed care population. Diabetes Care 29. Lee DH, Steffes MW, Jacobs DR Jr. analysis. 1997;20:1396–1402 Can persistent organic pollutants explain Prior Presentation. Parts of this study were 15. Committee opinion no. 504: screening and the association between serum gamma- presented in abstract form at the 7th Interna- diagnosis of gestational diabetes mellitus. Ob- glutamyltransferase and type 2 diabetes? tional Symposium on Diabetes, Hypertension, stet Gynecol 2011;118:751–753 Diabetologia 2008;51:402–407 Metabolic Syndrome, and Pregnancy, Florence, 16. Matthews DR, Hosker JP, Rudenski AS, 30. Nathwani RA, Pais S, Reynolds TB, Kaplowitz N. Italy, 14–16 March 2013. Naylor BA, Treacher DF, Turner RC. Homeostasis Serum alanine aminotransferase in skeletal muscle model assessment: insulin resistance and beta- diseases. Hepatology 2005;41:380–382 References cell function from fasting plasma glucose and 31. Tiikkainen M, Bergholm R, Vehkavaara S, 1. Metzger BE. Summary and recommenda- insulin concentrations in man. Diabetologia et al. Effects of identical weight loss on body tions of the Third International Workshop- 1985;28:412–419 composition and features of insulin resistance Conference on Gestational Diabetes Mellitus. 17. Doi Y, Kubo M, Yonemoto K, et al. Liver in obese women with high and low liver fat con- Diabetes 1991;40(Suppl. 2):197–201 enzymes as a predictor for incident diabetes tent. Diabetes 2003;52:701–707 2. American Diabetes Association. Gestational in a Japanese population: the Hisayama study. 32. Nannipieri M, Gonzales C, Baldi S, et al.; diabetes mellitus. Diabetes Care 2000;23(Suppl. Obesity (Silver Spring) 2007;15:1841–1850 Mexico City Diabetes Study. Liver enzymes, 1):S77–S79 18. Perry IJ, Wannamethee SG, Shaper AG. the metabolic syndrome, and incident diabetes: 3. Ferrara A, Kahn HS, Quesenberry CP, Riley C, Prospective study of serum gamma- the Mexico City diabetes study. Diabetes Care Hedderson MM. An increase in the incidence of glutamyltransferase and risk of NIDDM. Diabe- 2005;28:1757–1762 gestational diabetes mellitus: Northern California, tes Care 1998;21:732–737 33. Schindhelm RK, Dekker JM, Nijpels G, Heine 1991-2000. Obstet Gynecol 2004;103:526–533 19. Lee DH, Ha MH, Kim JH, et al. Gamma- RJ, Diamant M. No independent association of 4. Clark JM, Diehl AM. Defining nonalcoholic glutamyltransferase and diabetesda4year alanine aminotransferase with risk of future fatty liver disease: implications for epidemiologic follow-up study. Diabetologia 2003;46:359–364 type 2 diabetes in the Hoorn study. Diabetes studies. Gastroenterology 2003;124:248–250 20. Lee DH, Silventoinen K, Jacobs DR Jr, Jousilahti Care 2005;28:2812 5. Kugelman A, Choy HA, Liu R, Shi MM, Gozal E, P, Tuomileto J. gamma-Glutamyltransferase, obe- 34. Kunutsor SK, Apekey TA, Walley J. Liver ami- Forman HJ. gamma-Glutamyl transpeptidase is sity, and the risk of type 2 diabetes: observa- notransferases and risk of incident type 2 dia- increased by oxidative stress in rat alveolar L2 tional cohort study among 20,158 middle-aged betes: a systematic review and meta-analysis. epithelial cells. Am J Respir Cell Mol Biol 1994; men and women. J Clin Endocrinol Metab 2004; Am J Epidemiol 2013;178:159–171 11:586–592 89:5410–5414 35. Bozkurt L, Gobl¨ CS, Tura A, et al. Fatty liver 6. Lieberman MW, Barrios R, Carter BZ, et al. 21. Lee DH, Jacobs DR Jr, Gross M, et al. index predicts further metabolic deteriorations gamma-Glutamyl transpeptidase. What does the Gamma-glutamyltransferase is a predictor of in- in women with previous gestational diabetes. organization and expression of a multipromoter cident diabetes and hypertension: the Coronary PLoS One 2012;7:e32710