Since the 1960s, the organic compound bisphenol A has. Original Research

Transcription

1 Original Research Annals of Internal Medicine Relationship of Urinary Bisphenol A Concentration to Risk for Prevalent Type 2 Diabetes in Chinese Adults A Cross-sectional Analysis Guang Ning, MD, PhD; Yufang Bi, MD; Tiange Wang, MD; Min Xu, MD; Yu Xu, MD; Yun Huang, MD; Mian Li, MD; Xiaoying Li, MD, PhD; Weiqing Wang, MD; Yuhong Chen, MD; Yaohua Wu, MM; Jianing Hou, MD; Aiyun Song, MD; Yu Liu, MD; and Shenghan Lai, MD, MPH Background: Greater bisphenol A exposure has been shown to be associated with a higher risk for self-reported adverse health outcomes, including diabetes. Objective: To examine the association between bisphenol A exposure and type 2 diabetes in adults. Design: Cross-sectional study. Setting: Songnan, Baoshan District, Shanghai, China. Participants: 3423 local residents aged 40 years or older who were enrolled from 27 June 2008 to 10 August Measurements: Urinary concentrations of bisphenol A from morning spot urine samples (exposure) and fasting plasma glucose concentration, plasma glucose concentration 2 hours after an oral glucose tolerance test, and serum insulin concentration (outcomes). Results: Median age of the participants was 59.0 years (interquartile range, 53.0 to 68.7 years), 40% were men, and 1087 had type 2 diabetes. The median urinary bisphenol A level was 0.81 ng/ml (interquartile range, 0.47 to 1.43 ng/ml). Clinical characteristics differed between participants with normal glucose regulation and those with impaired glucose regulation and by bisphenol A quartile, but in multivariable analyses, there was no clear association between bisphenol A levels and type 2 diabetes. The adjusted odds ratio (OR) of type 2 diabetes was slightly increased for participants in the second bisphenol A quartile (0.48 to 0.81 ng/ml) (adjusted OR, 1.30 [95% CI, 1.03 to 1.64]) and the fourth quartile ( 1.43 ng/ml) (adjusted OR, 1.37 [CI, 1.08 to 1.74]) but not the third quartile (0.82 to 1.43 ng/ml) (adjusted OR, 1.09 [CI, 0.86 to 1.39]), and a test of the trend of the association was not statistically significant. Limitations: The cross-sectional study design and nonrandom sample of participants limit the conclusions that can be drawn. Many patients in the study already had diabetes, successful treatment of which could have obscured apparent associations. Dietary variables were not measured; however, this is necessary in observational studies of bisphenol A and diabetes because the presence of the chemical in the body may reflect consumption of sugared drinks in plastic bottles. Conclusion: These findings do not confirm a previously reported association between urinary bisphenol A levels and self-reported type 2 diabetes. Primary Funding Source: Key Laboratory for Endocrine and Metabolic Diseases of Ministry of Health, National Key New Drug Creation and Manufacturing Program of Ministry of Science and Technology, Sector Funds of Ministry of Health, Creative Research Group of Ministry of Education, Major Project of Shanghai Committee of Science and Technology, National Key Technologies Research and Development Program of Ministry of Science and Technology, and the National Natural Science Foundation of China. Ann Intern Med. 2011;155: For author affiliations, see end of text. Since the 1960s, the organic compound bisphenol A has been present in many hard plastic containers, including reusable water bottles, baby bottles, and metal-based food and beverage cans (1). According to studies from the Centers for Disease Control and Prevention, bisphenol A was found in the urine of 95% of adults sampled in 1988 to 1994 and in 93% of children and adults tested in 2003 to 2004 (2, 3). Animal studies have suggested that exposure See also: Print Editors Notes Editorial comment Summary for Patients....I-36 Web-Only Appendix Conversion of graphics into slides to bisphenol A may contribute to or exacerbate the development of type 2 diabetes (4 7), but epidemiologic data in humans were not available until recently. A recent study of data from NHANES (National Health and Nutrition Examination Survey) concluded that higher urinary bisphenol A concentrations may be associated with self-reported adverse health outcomes, including diabetes (8). We sought to confirm the association between bisphenol A and diabetes in a community-based investigation in China. METHODS Study Design and Participants We designed a cross-sectional study to investigate the association between bisphenol A exposure and type 2 diabetes in Chinese adults. We enrolled study participants from Songnan Community, Baoshan District, Shanghai, China, in 2 phases. In phase 1 (June and July 2008), all registered permanent residents aged 40 years or older were American College of Physicians

2 Bisphenol A and Diabetes Original Research invited to receive a screening examination; persons participated. We tested fasting plasma glucose and preliminarily categorized participants into 3 groups: normal glucose regulation (NGR), defined as a fasting plasma glucose level less than 5.6 mmol/l ( 101 mg/dl) and no history of diabetes; impaired glucose regulation (IGR), defined as a fasting plasma glucose level of 5.6 to less than 7.0 mmol/l (101 to 126 mg/dl) and no history of diabetes; and diabetes, defined as a fasting plasma glucose level of 7.0 mmol/l or greater ( 126 mg/dl) or a history of diabetes. In phase 2 (June through August 2009), we randomly selected participants from the 3 groups in a ratio of 1.0 (diabetes) to 1.2 (IGR) to 1.44 (NGR), oversampling people with lower glucose levels because they might have a lower participation rate than those with higher glucose levels. The selected participants received a comprehensive examination that included a detailed questionnaire, anthropometric measurement, a standard 75-g oral glucose tolerance test, and blood and urine collection. Among 3455 persons with blood and urine samples included in the phase 2 survey, 32 were excluded from the current analysis because results of the oral glucose tolerance test at 0 or 2 hours were missing. Thus, 3423 participants were included in the final analysis. The 3423 participants and the 6762 nonparticipants were similar in terms of such characteristics as age, sex, body mass index (BMI), and blood pressure. Measurements Interviews about sociodemographic characteristics, medical history, family history, and lifestyle factors were conducted by trained personnel. Clinical examinations, including measurements of weight, height, waist circumference, and blood pressure, were performed by experienced nurses according to a standard protocol. Three sitting blood pressure measurements that were obtained consecutively at 1-minute intervals by using an automated electronic device (OMRON Model HEM-752 FUZZY, Omron Company, Dalian, China) were averaged for analysis. All participants underwent a 75-g oral glucose tolerance test after an overnight fast of more than 10 hours, and plasma glucose and serum insulin concentrations were obtained at 0 and 2 hours during the test. Laboratory tests included measurement of fasting plasma glucose, plasma glucose 2 hours after an oral glucose tolerance test, serum insulin, serum alanine aminotransferase (ALT), -glutamyltransferase (GGT), serum albumin, total bilirubin, serum creatinine, lipid profile (total cholesterol, highdensity lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides), and high sensitivity C-reactive protein. The glucose and insulin samples were centrifuged on site at 4 C and measured within 1 hour. Plasma or serum was immediately moved into Eppendorf tubes and stored at 80 C. Blood glucose was measured by using the glucose oxidase method on an autoanalyzer (ADVIA-1650 Context Bisphenol A is a chemical in plastic that some evidence suggests may contribute to chronic disease, including type 2 diabetes. Contribution In this cross-sectional study from China, researchers were unable to detect a clear association between urinary bisphenol A concentrations and objective measures of impaired glucose regulation, including type 2 diabetes. Caution The study did not assess diet or treatment of existing diabetes. Implication This study could not confirm a previously reported association between bisphenol A and type 2 diabetes. Larger cohort studies accounting for diet and incident disease are needed to draw more definitive conclusions about the safety of bisphenol A. The Editors Chemistry System, Bayer, Leverkusen, Germany), and serum insulin was measured by using an electrochemiluminescence assay (Roche Diagnostics, Basel, Switzerland). Measurement of ALT, GGT, albumin, bilirubin, creatinine, triglycerides, cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and high sensitivity C-reactive protein was done with an autoanalyzer (ADVIA-1650 Chemistry System). The abbreviated Modification of Diet in Renal Disease Study formula as recalibrated for Chinese persons (9) was used to calculate the estimated glomerular filtration rate: 186 (serum creatinine value in mol/l 0.011) (age) (0.742 if female) (the number is the adjusting coefficient for Chinese persons). Type 2 diabetes was defined as 1) a fasting plasma glucose concentration of 7.0 mmol/l or more ( 126 mg/dl) or a plasma glucose concentration of 11.1 mmol/l or more ( 200 mg/dl) 2 hours after an oral glucose tolerance test (10), or 2) receipt of therapy for diabetes. To measure urinary bisphenol A concentration, a morning spot urine sample was collected from each study participant. The Committee on Human Research at Shanghai Jiao-Tong University School of Medicine Rui-Jin Hospital approved the study protocol, and all study participants provided written informed consent. All procedures used in this study were in accordance with institutional guidelines. Quantification of Urinary Bisphenol A Concentration Total free and conjugated urinary bisphenol A was measured at the Shanghai Institute of Materia Medica, Chinese Academy of Sciences, and determined after enzymatic hydrolysis by a sensitive and selective liquid 20 September 2011 Annals of Internal Medicine Volume 155 Number 6 369

3 Original Research Bisphenol A and Diabetes Table 1. Participant Characteristics* Characteristic Overall Glucose Metabolism Status NGR IGR Type 2 Diabetes Participants, n Age, y 59.0 ( ) 56.8 ( ) 59.8 ( ) 62.7 ( ) Men, n (%) 1369 (40.0) 576 (36.3) 293 (39.2) 500 (46.0) Family history of diabetes, n (%) 651 (19.0) 215 (13.5) 124 (16.6) 312 (28.7) Educational level, n (%) 6 y 764 (22.8) 257 (16.5) 170 (23.4) 337 (32.6) y 1651 (49.3) 832 (53.5) 346 (47.7) 473 (44.3) 9 y 935 (27.9) 467 (30.0) 210 (28.9) 258 (24.2) Smoking status, n (%) Never-smoker 2538 (74.2) 1180 (74.3) 564 (75.4) 794 (73.1) Former smoker 179 (5.2) 67 (4.2) 41 (5.5) 71 (6.5) Current smoker 706 (20.6) 341 (21.5) 143 (19.1) 222 (20.4) Alcohol consumption, n (%) Never-drinker 2849 (83.2) 1336 (84.1) 611 (81.7) 902 (83.0) Former drinker 39 (1.1) 15 (0.9) 10 (1.3) 14 (1.3) Current drinker 535 (15.6) 237 (14.9) 127 (17.0) 171 (15.7) Body mass index, kg/m ( ) 24.3 ( ) 25.4 ( ) 26.0 ( ) Waist circumference, cm 87.0 ( ) 84.0 ( ) 88.5 ( ) 90.0 ( ) Blood pressure, mm Hg Systolic 137 ( ) 129 ( ) 140 ( ) 146 ( ) Diastolic 78 (71 85) 77 (70 84) 80 (73 86) 80 (73 87) Total cholesterol level mmol/l 5.12 ( ) 5.03 ( ) 5.19 ( ) 5.20 ( ) mg/dl 198 ( ) 194 ( ) 200 ( ) 201 ( ) LDL cholesterol level mmol/l 2.36 ( ) 2.32 ( ) 2.45 ( ) 2.38 ( ) mg/dl 91 (75 109) 90 (74 107) 96 (78 113) 92 (74 112) HDL cholesterol level mmol/l 1.32 ( ) 1.36 ( ) 1.33 ( ) 1.28 ( ) mg/dl 52 (44 59) 53 (46 61) 51 (44 59) 49 (43 56) Triglyceride level mmol/l 1.44 ( ) 1.23 ( ) 1.49 ( ) 1.70 ( ) mg/dl 127 (86 184) 109 (80 158) 132 (97 194) 150 ( ) Fasting plasma glucose level mmol/l 5.2 ( ) 4.9 ( ) 5.2 ( ) 6.9 ( ) mg/dl 94 (87 108) 88 (83 94) 94 (87 103) 124 ( ) 2-h oral glucose tolerance test plasma glucose level mmol/l 7.9 ( ) 6.1 ( ) 8.8 ( ) 15.9 ( ) mg/dl 142 ( ) 110 (95 124) 159 ( ) 286 ( ) Fasting insulin level, IU/mL 7.21 ( ) 5.97 ( ) 8.03 ( ) 9.15 ( ) hs-crp level, mg/l 0.20 ( ) 0.12 ( ) 0.24 ( ) 0.40 ( ) ALT level, U/L 19 (13 27) 16 (12 24) 20 (14 29) 22 (15 34) GGT level, U/L 24 (17 36) 20 (15 30) 25 (18 41) 29 (21 44) Total bilirubin level, mol/l 13 (11 17) 13 (11 16) 13 (11 17) 14 (11 18) mg/dl 0.76 ( ) 0.76 ( ) 0.76 ( ) 0.89 ( ) Albumin level, g/l 44 (42 45) 43 (42 45) 44 (42 45) 44 (43 45) Serum creatinine level mol/l 69 (60 82) 69 (60 80) 71 (60 82) 69 (59 83) mg/dl 0.8 ( ) 0.8 ( ) 0.8 ( ) 0.8 ( ) Estimated glomerular filtration rate, ml/min per 1.73 m ( ) ( ) ( ) ( ) Type 2 diabetes, n (%) 1087 (31.8) Bisphenol A level, ng/ml 0.81 ( ) 0.79 ( ) 0.79 ( ) 0.82 ( ) ALT alanine aminotransferase; GGT -glutamyltransferase; HDL high-density lipoprotein; hs-crp high-sensitivity C-reactive protein; IGR impaired glucose regulation; LDL low-density lipoprotein; NGR normal glucose regulation. * Unless otherwise indicated, data are medians (interquartile ranges). Adjusted for age and sex. chromatography tandem mass spectrometry method (Appendix, available at Statistical Analysis Statistical analysis was performed by using SAS software, version 9.2 (SAS Institute, Cary, North Carolina). All continuous variables are summarized as medians and interquartile ranges, and all categorical variables as proportions. To compare between-group differences in demographic and clinical characteristics, laboratory variables, and other factors, we used the nonparametric Wilcoxon September 2011 Annals of Internal Medicine Volume 155 Number 6

4 Bisphenol A and Diabetes Original Research Table 1 Continued P Value for Trend Bisphenol A Quartile P Value for Trend 1(<0.47 ng/ml) 2 ( ng/ml) 3 ( ng/ml) 4 (>1.43 ng/ml) ( ) 59.9 ( ) 57.6 ( ) 57.1 ( ) 263 (31.0) 354 (40.7) 359 (42.2) 393 (46.0) (17.3) 167 (19.2) 168 (19.8) 169 (19.8) (30.0) 199 (23.2) 162 (19.5) 154 (18.5) 375 (45.2) 433 (50.5) 418 (50.4) 425 (51.0) 206 (24.8) 226 (26.3) 249 (30.0) 254 (30.5) (81.1) 644 (74.0) 624 (73.4) 582 (68.1) 45 (5.3) 45 (5.2) 44 (5.2) 45 (5.3) 115 (13.6) 181 (20.8) 182 (21.4) 228 (26.7) (88.2) 718 (82.5) 710 (83.5) 673 (78.7) 8 (0.9) 9 (1.0) 10 (1.2) 12 (1.4) 92 (10.9) 143 (16.4) 130 (15.3) 170 (19.9) ( ) 25.1 ( ) 25.1 ( ) 25.3 ( ) ( ) 88.0 ( ) 87.0 ( ) 87.1 ( ) ( ) 137 ( ) 135 ( ) 134 ( ) (70 85) 77 (71 85) 79 (72 85) 79 (72 85) ( ) 5.13 ( ) 5.11 ( ) 5.07 ( ) 199 ( ) 197 ( ) 197 ( ) 196 ( ) ( ) 2.34 ( ) 2.39 ( ) 2.37 ( ) 91 (75 110) 90 (75 109) 92 (74 109) 92 (76 108) ( ) 1.34 ( ) 1.31 ( ) 1.30 ( ) 52 (46 61) 52 (44 59) 51 (44 59) 50 (44 58) ( ) 1.43 ( ) 1.44 ( ) 1.45 ( ) 127 (89 182) 127 (87 186) 127 (86 186) 128 (86 184) ( ) 5.2 ( ) 5.2 ( ) 5.2 ( ) 94 (84 106) 94 (86 112) 94 (86 108) 94 (86 112) ( ) 8.0 ( ) 7.6 ( ) 8.0 ( ) 144 ( ) 144 ( ) 137 ( ) 144 ( ) ( ) 7.06 ( ) 6.89 ( ) 7.16 ( ) ( ) 0.21 ( ) 0.19 ( ) 0.22 ( ) (13 26) 19 (13 28) 18 (13 27) 20 (14 29) (16 33) 24 (17 37) 24 (17 36) 25 (18 39) (11 17) 14 (11 17) 13 (11 17) 14 (11 17) 0.76 ( ) 0.89 ( ) 0.76 ( ) 0.89 ( ) (42 45) 43 (42 45) 44 (42 45) 44 (42 45) (59 81) 70 (60 81) 69 (59 82) 70 (60 81) 0.8 ( ) 0.8 ( ) 0.8 ( ) 0.8 ( ) ( ) ( ) ( ) ( ) (30.3) 296 (34.0) 250 (29.4) 284 (33.2) ( ) 0.63 ( ) 1.05 ( ) 2.27 ( ) 2-sample test for continuous variables and the Fisher exact test for categorical variables. We first studied the relation between urinary bisphenol A levels and the presence of diabetes according to quartiles of urinary bisphenol A levels. We computed odds ratios with the lowest quartile as the reference group. Univariate logistic regression models were first fitted to evaluate the crude association between the presence of type 2 diabetes and each individual risk factor: age; sex; educational attainment; alcohol consumption; cigarette smoking; family history of diabetes; BMI; waist circumference; lipid profile (total cholesterol, high-density lipoprotein choles September 2011 Annals of Internal Medicine Volume 155 Number 6 371

5 Original Research Bisphenol A and Diabetes Table 2. Odds Ratios for the Association Between Urinary Bisphenol A Level and Risk for Type 2 Diabetes* Characteristic Univariate Model Odds Ratio (95% CI) Final Model Age 1.04 ( ) 1.05 ( ) Female sex 0.70 ( ) 0.69 ( ) Educational level 6 y 1.00 (reference) 1.00 (reference) y 0.51 ( ) 0.65 ( ) 9 y 0.48 ( ) 0.62 ( ) Cigarette smoking Never-smoker 1.00 (reference) Former smoker 0.76 ( ) Ever-smoker 1.09 ( ) Alcohol consumption Never consumed alcohol 1.00 (reference) Ever consumed alcohol 1.03 ( ) Positive family history of 2.37 ( ) 3.25 ( ) diabetes Body mass index 1.11 ( ) Waist circumference 1.05 ( ) 1.05 ( ) Systolic blood pressure 1.03 ( ) 1.01 ( ) LDL cholesterol level 1.08 ( ) HDL cholesterol level 0.41 ( ) Total cholesterol level 1.10 ( ) ln(triglyceride level) 2.17 ( ) 1.59 ( ) ln(hs-crp level) 1.30 ( ) 1.19 ( ) ln(alt level) 2.13 ( ) 1.60 ( ) ln(ggt level) 2.04 ( ) Estimated glomerular 1.00 ( ) 1.01 ( ) filtration rate Albumin level 1.11 ( ) 1.13 ( ) Total bilirubin level 1.03 ( ) 1.03 ( ) Bisphenol A quartile 1( 0.47 ng/ml) 1.00 (reference) 1.00 (reference) 2 ( ng/ml) 1.19 ( ) 1.30 ( ) 3 ( ng/ml) 0.96 ( ) 1.09 ( ) 4( 1.43 ng/ml) 1.14 ( ) 1.37 ( ) ALT alanine aminotransferase; GGT -glutamyltransferase; HDL highdensity lipoprotein; hs-crp high-sensitivity C-reactive protein; LDL lowdensity lipoprotein. * Total of 3423 participants, of whom 1087 had new or previously diagnosed diabetes at baseline. Adjusted for age, sex, educational level, family history of diabetes, waist circumference, systolic blood pressure, ln(triglyceride level), ln(hs-crp level), ln(alt level), estimated glomerular filtration rate, albumin level, and total bilirubin level. The final model was based on a generalized additive analysis, in which age and waist circumference were modeled nonparametrically with smoothing splines. The odds ratios for age and waist circumference were estimated for the linear parametric components. terol, low-density lipoprotein cholesterol, and triglyceride levels), high-sensitivity C-reactive protein level; blood pressure; levels of albumin, total bilirubin, ALT, GGT; and estimated glomerular filtration rate. Factors that were significant at the P 0.20 level in the univariate models were put into the multiple logistic regression models to identify the ones independently associated with the presence of type 2 diabetes. Then, the importance of each variable included in this model was evaluated with an examination of the Wald statistic for each variable in the model and a comparison of each estimated regression coefficient in the multivariate model with the regression coefficient from the corresponding univariate model. Variables that did not make significant contributions to the models on the basis of these 2 criteria were deleted in a stagewise manner, and a new model was refitted. This process of eliminating, refitting, and verifying continued until all of the variables included were statistically significant, yielding a final model (11). Goodness-of-fit of the final model was examined by using the Hosmer Lemeshow goodness-of-fit test (11). Because some potential confounding factors in the final logistic regression model may be powerful, the effects of waist circumference and age were modelled nonparametrically by using smoothing splines in a generalized additive model (the results of the final model in Table 2 were derived from the generalized additive model) (12). Scatterplots showing the smoothed independent variable values plotted against the partial residuals were used to evaluate the relationship between the independent variable with the residualized (adjusted) values for dependent variables (12, 13). Because some laboratory variables were not normally distributed, log-normal transformation was implemented before those data were entered into the logistic regression models. The reported P values were 2-sided, and P 0.05 was considered statistically significant. Role of the Funding Source This study was funded by grants from Key Laboratory for Endocrine and Metabolic Diseases of Ministry of Health, National Key New Drug Creation and Manufacturing Program of Ministry of Science and Technology, Sector Funds of Ministry of Health, Creative Research Group of Ministry of Education, Major Project of Shanghai Committee of Science and Technology, National Key Technologies Research and Development Program of Ministry of Science and Technology, and the National Natural Science Foundation of China. The sponsors of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study. The management committee had final responsibility for the decision to submit the manuscript for publication. RESULTS Table 1 shows the general characteristics of study participants. Overall, median age was 59.0 years (interquartile range [IQR], 53.0 to 68.7 years), 40% were men, 1087 had type 2 diabetes (688 with a previously diagnosis and 399 with a new diagnosis), 748 had IGR, and 1588 had NGR. The median urinary bisphenol A level was 0.81 ng/ml (IQR, 0.47 to 1.43 ng/ml). Participants with type 2 diabetes differed from those with NGR or IGR in multiple clinical variables. Although the median urinary bisphenol A concentrations were similar in the NGR (0.79 ng/ml [IQR, 0.76 to 0.84 ng/ml]), IGR (0.79 ng/ml [IQR, 0.74 to 0.85 ng/ml]), and type 2 diabetes (0.82 ng/ml [IQR, 0.78 to 0.87 ng/ml]) groups, September 2011 Annals of Internal Medicine Volume 155 Number 6

6 Bisphenol A and Diabetes Original Research the slightly higher concentration in patients with type 2 diabetes was statistically significant after adjustment for age and sex (P 0.01). Many clinical variables also differed among bisphenol A quartiles: There were more younger participants and men in the highest quartile. Across quartiles, BMI, waist circumference, and fasting glucose levels and plasma glucose levels 2 hours after an oral glucose tolerance test gradually increased and high-density lipoprotein cholesterol level decreased. Higher proportions of participants with type 2 diabetes were found in the second and highest bisphenol A quartile compared with the lowest. Multivariable analysis did not reveal a clear association between bisphenol A concentration and type 2 diabetes. After adjustment for the variables listed in Table 2, the odds of having type 2 diabetes was slightly increased for participants with urinary bisphenol A concentrations in the second quartile (0.48 to 0.81 ng/ml) (adjusted odds ratio, 1.30 [95% CI, 1.03 to 1.64]) and fourth quartile ( 1.43 ng/ml) (adjusted odds ratio, 1.37 [CI, 1.08 to 1.74]) but not in the third quartile (0.82 to 1.43 ng/ml) (adjusted odds ratio, 1.09 [CI, 0.86 to 1.39]), and a test of the trend of the association was not statistically significant. DISCUSSION Bisphenol A has been widely used in plastics and the coatings of metal cans for decades, leading to broad and worldwide exposure to this organic compound. Bisphenol A has a fast metabolic rate, with a half-life less than 6 hours after oral administration, and urine is considered the body fluid most appropriate for assessment of bisphenol A exposure (14). Such exposure has been reported in approximately 95% of the U.S. population on the basis of detection in urine (3). Animal studies have suggested that bisphenol A is an endocrine disruptor that may contribute to or exacerbate development of type 2 diabetes (4 7, 14 16). In vivo, bisphenol A acts as a metabolic disturber to cause adverse human health effects by damaging liver function, disrupting pancreatic -cell function, disturbing thyroid hormone, and promoting obesity. An animal study showed that even at doses much lower than the U.S. Environmental Protection Agency s lowest observed level of adverse effects, bisphenol A induces insulin resistance (17). Bisphenol A also has been demonstrated to enhance insulin biosynthesis and secretion in -cells in adult male mice, contributing to -cell overstimulation and dysfunction (5). Bisphenol A at environmentally relevant doses also could suppress adiponectin release from human adipocytes (18). Previous animal studies have shed light on the potential mechanisms by which bisphenol A disturbs metabolism mainly by binding to a membrane estrogenic receptor. As a consequence, pancreatic -cells are overstimulated, inducing postprandial hyperinsulinemia, glucose intolerance, and insulin resistance. Evidence that bisphenol A may be an endocrine disruptor in animals has raised concerns that bisphenol A might pose a health threat to humans. Nevertheless, no large-scale human epidemiologic studies have been published. Only 1 published study, by Lang and colleagues (8), has reported positive findings based on secondary analyses of data in humans. To our knowledge, our cross-sectional study is the largest population-based study of bisphenol A to date and the only one to focus solely on the association with type 2 diabetes. We did not find a clear association between bisphenol A and risk for diabetes, nor did we find a clear, monotonic association between bisphenol A and risk for diabetes. Our study differs from that of Lang and colleagues (8) in 3 key respects. First, in the latter study, the diagnoses of diabetes were self-reported, and borderline diabetes cases and self-reported diagnoses of diabetes were combined into a single group. In contrast, we conducted a 75-g oral glucose tolerance test to classify glucose metabolism status and diagnose type 2 diabetes, making it possible to observe changes in bisphenol A concentration in groups ranging from NGR to IGR to type 2 diabetes. Second, our study population was much larger than that of Lang and colleagues, which had 1455 participants a difference that could minimize the bias in epidemiologic analysis. Finally, Lang and colleagues finding that a 1-SD increase in bisphenol A concentration was associated with an increased odds ratio of reporting diabetes (1.39 [CI, 1.21 to 1.60] was based on a hidden assumption that the relationship between the bisphenol A level and the risk for diabetes was linear or monotonic. However, we could not find such a linear relationship in either our data or those of NHANES Our study has several limitations. First, the study participants were not a random sample of the people in the community, and the results should therefore be interpreted with caution. Second, because our study is cross-sectional, no causal conclusions can be made, even about its negative findings. Third, our analyses are based on morning urinary concentrations of bisphenol A, which reflect only recent exposure. Fourth, 599 of 1087 persons with diabetes were receiving oral antihyperglycemics or insulin (or both). These treatments may have decreased both fasting plasma glucose levels and those obtained 2 hours after an oral glucose tolerance test and could have diluted the association between glucose levels and bisphenol A. Finally, we did not collect data on socioeconomic status, diet, dietary behaviors, and physical activity. Bisphenol A could be a marker of consumption of sugared drinks (such as soda) in plastic bottles, and intake of these beverages has been shown to be associated with glucose dysregulation. In the absence of reliable dietary measures, any observational study that detects an association could actually be detecting 20 September 2011 Annals of Internal Medicine Volume 155 Number 6 373

7 Original Research Bisphenol A and Diabetes the association between higher exposure to sugared drinks and type 2 diabetes. In summary, our data do not support the previous finding of an association between urinary bisphenol A levels and self-reported type 2 diabetes (8). Although we used fasting plasma glucose levels and oral glucose tolerance testing to define whether participants had diabetes, our study is limited by its cross-sectional design. We are continuing to gather data prospectively about type 2 diabetes incidence in the same community in a study that will provide a higher level of evidence to support or refute the previously reported association between bisphenol A and type 2 diabetes. From Key Laboratory for Endocrine and Metabolic Diseases of Ministry of Health, Rui-Jin Hospital, Shanghai Jiao-Tong University School of Medicine, E-Institute of Shanghai Universities, Shanghai, China; Shanghai Clinical Center for Endocrine and Metabolic Diseases, Shanghai Institute of Endocrine and Metabolic Diseases, Department of Endocrinology and Metabolism, Rui-Jin Hospital, Shanghai Jiao-Tong University School of Medicine, Shanghai, China; and Johns Hopkins University School of Medicine, Baltimore, Maryland. Acknowledgment: The authors thank the field workers for their contribution and the participants for their cooperation. Grant Support: By Key Laboratory for Endocrine and Metabolic Diseases of Ministry of Health (1994DP131044), National Key New Drug Creation and Manufacturing Program of Ministry of Science and Technology (2008ZX ), Sector Funds of Ministry of Health ( ), Creative Research Group of Ministry of Education (IRT0932), Major Project of Shanghai Committee of Science and Technology (09DZ ), National Key Technologies Research and Development Program of Ministry of Science and Technology (2008BAI52B03), and the National Natural Science Foundation of China ( ). Potential Conflicts of Interest: Disclosures can be viewed at M Reproducible Research Statement: Study protocol and data set: Not available. Statistical code: Available from Dr. Ning ( , gning@sibs.ac.cn). Corresponding Author: Guang Ning, MD, PhD, Shanghai Clinical Center for Endocrine and Metabolic Diseases, Shanghai Institute of Endocrine and Metabolic Diseases, Department of Endocrinology and Metabolism, Rui-Jin Hospital, Shanghai Jiao-Tong University School of Medicine, 197 Rui-Jin 2nd Road, Shanghai , China; , gning@sibs.ac.cn. Current author addresses and author contributions are available at References 1. U.S. Food and Drug Administration. Update on bisphenol A (BPA) for use in food: January Accessed at /ucm htm on 1 June Calafat AM, Kuklenyik Z, Reidy JA, Caudill SP, Ekong J, Needham LL. Urinary concentrations of bisphenol A and 4-nonylphenol in a human reference population. Environ Health Perspect. 2005;113: [PMID: ] 3. Calafat AM, Ye X, Wong LY, Reidy JA, Needham LL. Exposure of the U.S. population to bisphenol A and 4-tertiary-octylphenol: Environ Health Perspect. 2008;116: [PMID: ] 4. Nadal A, Alonso-Magdalena P, Soriano S, Quesada I, Ropero AB. The pancreatic beta-cell as a target of estrogens and xenoestrogens: implications for blood glucose homeostasis and diabetes. Mol Cell Endocrinol. 2009;304:63-8. [PMID: ] 5. Alonso-Magdalena P, Ropero AB, Carrera MP, Cederroth CR, Baquié M, Gauthier BR, et al. Pancreatic insulin content regulation by the estrogen receptor ER alpha. PLoS One. 2008;3:e2069. [PMID: ] 6. Ropero AB, Alonso-Magdalena P, García-García E, Ripoll C, Fuentes E, Nadal A. Bisphenol-A disruption of the endocrine pancreas and blood glucose homeostasis. Int J Androl. 2008;31: [PMID: ] 7. Alonso-Magdalena P, Laribi O, Ropero AB, Fuentes E, Ripoll C, Soria B, et al. Low doses of bisphenol A and diethylstilbestrol impair Ca2 signals in pancreatic alpha-cells through a nonclassical membrane estrogen receptor within intact islets of Langerhans. Environ Health Perspect. 2005;113: [PMID: ] 8. Lang IA, Galloway TS, Scarlett A, Henley WE, Depledge M, Wallace RB, et al. Association of urinary bisphenol A concentration with medical disorders and laboratory abnormalities in adults. JAMA. 2008;300: [PMID: ] 9. Ma YC, Zuo L, Chen JH, Luo Q, Yu XQ, Li Y, et al. Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease. J Am Soc Nephrol. 2006;17: [PMID: ] 10. Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med. 1998;15: [PMID: ] 11. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley & Sons; Hastie TJ, Tibshirani RJ. Generalized additive models (with discussion), Statistical Science. 1986;1: Rosenberg PS, Katki H, Swanson CA, Brown LM, Wacholder S, Hoover RN. Quantifying epidemiologic risk factors using non-parametric regression: model selection remains the greatest challenge. Stat Med. 2003;22: [PMID: ] 14. Dekant W, Völkel W. Human exposure to bisphenol A by biomonitoring: methods, results and assessment of environmental exposures. Toxicol Appl Pharmacol. 2008;228: [PMID: ] 15. Vogel SA. The politics of plastics: the making and unmaking of bisphenol a safety. Am J Public Health. 2009;99 Suppl 3:S [PMID: ] 16. Wetherill YB, Akingbemi BT, Kanno J, McLachlan JA, Nadal A, Sonnenschein C, et al. In vitro molecular mechanisms of bisphenol A action. Reprod Toxicol. 2007;24: [PMID: ] 17. Alonso-Magdalena P, Morimoto S, Ripoll C, Fuentes E, Nadal A. The estrogenic effect of bisphenol A disrupts pancreatic beta-cell function in vivo and induces insulin resistance. Environ Health Perspect. 2006;114: [PMID: ] 18. Hugo ER, Brandebourg TD, Woo JG, Loftus J, Alexander JW, Ben- Jonathan N. Bisphenol A at environmentally relevant doses inhibits adiponectin release from human adipose tissue explants and adipocytes. Environ Health Perspect. 2008;116: [PMID: ] September 2011 Annals of Internal Medicine Volume 155 Number 6

8 Annals of Internal Medicine Current Author Addresses: Drs. Ning, Bi, T. Wang, M. Xu, Y. Xu, Huang, M. Li, X. Li, W. Wang, Chen, Wu, Hou, Song, and Liu: Shanghai Clinical Center for Endocrine and Metabolic Diseases, Shanghai Institute of Endocrine and Metabolic Diseases, Department of Endocrinology and Metabolism, Rui-Jin Hospital, Shanghai Jiao-Tong University School of Medicine, 197 Rui-Jin 2nd Road, Shanghai , China. Dr. Lai: Johns Hopkins School of Medicine, 600 North Wolfe Street, Pathology 301, Baltimore, MD Author Contributions: Conception and design: G. Ning, S. Lai. Analysis and interpretation of the data: G. Ning, Y. Bi, T. Wang, M. Xu, Y. Xu, Y. Huang, S. Lai. Drafting of the article: G. Ning, Y. Bi, T. Wang, M. Xu, Y. Xu, Y. Huang, S. Lai. Critical revision of the article for important intellectual content: G. Ning, Y. Bi, T. Wang, M. Xu, Y. Xu, Y. Huang. Final approval of the article: G. Ning, Y. Bi, T. Wang, M. Xu, Y. Xu, Y. Huang, M. Li, X. Li, W. Wang, Y. Chen, Y. Wu, J. Hou, A. Song, Y. Liu, S. Lai. Provision of study materials or patients: G. Ning. Statistical expertise: G. Ning, Y. Bi, T. Wang, M. Xu, Y. Xu, Y. Huang, S. Lai. Obtaining of funding: G. Ning. Administrative, technical, or logistic support: G. Ning. Collection and assembly of data: Y. Bi, T. Wang, M. Xu, Y. Xu, Y. Huang, M. Li, X. Li, W. Wang, Y. Chen, Y. Wu, J. Hou, A. Song, Y. Liu. APPENDIX: QUANTIFICATION OF URINARY BISPHENOL A CONCENTRATION Total (free and conjugated) urinary bisphenol A concentrations were measured at the Shanghai Institute of Materia Medica, Chinese Academy of Sciences, and were determined after enzymatic hydrolysis by a sensitive and selective liquid chromatography tandem mass spectrometry method. In brief, 50 L of the internal standard (500 ng/ml bisphenol A-d6) and 300 L of -Dglucuronidase (2000 U/mL in 1 M citrate buffer, ph 5.0) were added to a 300- L aliquot of urine. The mixture was incubated in a water bath at 37 C overnight (16 hours). The samples were then extracted after the addition of 100 L of 1 M phosphoric acid and 3 ml of dichloromethane. The organic phase was separated, evaporated to dryness, reconstituted, and injected into the liquid chromatography tandem mass spectrometry system, which consisted of a Shimadzu (Kyoto, Japan) LC-10AD solvent delivery pump, an SIL-HT A autosampler (Shimadzu, Kyoto, Japan), and an Applied Biosystems (Concord, Ontario, Canada) API4000 triple quadrupole mass spectrometer fitted with a heated nebulizer interface. The mass spectrometer was operated in the negative ion detection mode. Multiple reaction monitoring included m/z m/z 133 for bisphenol A and m/z 2333 m/z 138 for bisphenol A-d6, with a dwell time of 0.20 second per transition. Chromatographic separation was performed on a Zorbax Extend C18 column (150 mm 4.6 mm internal diameter; 5- m particle size; Agilent, Wilmington, Delaware). The mobile phase was composed of methanol and 0.175% ammonia water (volume ratio, 65:35) at a flow rate of 0.8 ml/min. The standard curve ranged from 0.30 to 100 ng/ml for bisphenol A using weighted (1/x 2 ) least-squares linear regression mode. Both the intraday and interday relative SDs, calculated from quality control samples (0.80, 10.0, and 80.0 ng/ml), were less than 8.7%. The interday relative error as determined from quality control samples was within 2.5%. The lower limit of quantification for bisphenol A, at which both precision and accuracy were less than 9.1%, was 0.30 ng/ml. For bisphenol A levels less than 0.30 ng/ml, a value of 0.15 ng/ml was assigned for the purpose of analysis. The results of stability experiments showed that no substantial degradation occurred during sample incubation, chromatography, or extraction processes for bisphenol A in urine samples (deviating no more than 7.4% at low and high nominal concentrations). The analytical data demonstrated excellent reproducibility, specificity, sensitivity, and precision for measurement of urinary bisphenol A concentrations September 2011 Annals of Internal Medicine Volume 155 Number 6 W-115