American Journal of Epidemiology The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. Vol. 178, No. 7 DOI: 10.1093/aje/kwt088 Advance Access publication: July 25, 2013 Original Contribution Prevalence of Diagnosed and Undiagnosed Type 2 Diabetes Mellitus Among US Adolescents: Results From the Continuous NHANES, 1999-2010 Ryan T. Demmer*, Aleksandra M. Zuk, Michael Rosenbaum, and Mose Desvarieux * Correspondence to Dr. Ryan T. Demmer, Department of Epidemiology, Mailman School of Public Health, Columbia University, 722 West 168th Street, New York, NY 10040 (e-mail: rtd2106@columbia.edu). Although prevalence and incidence of type 2 diabetes mellitus (T2DM) are reportedly increasing among adolescents, national data are lacking, particularly in regard to undiagnosed T2DM. To estimate the prevalence of diagnosed and undiagnosed T2DM among US adolescents, we analyzed a nationally representative cross-section of 11,888 adolescents aged 12-19 years who received a diabetes interview in the Continuous National Health and Nutrition Examination Survey during 1999-2010. Among them, a random subsample of 4,661 adolescents also had fasting blood samples collected. Persons who reported a previous diabetes diagnosis and were either taking no medication or taking an oral hypoglycemic agent (with or without insulin) were classified as having T2DM; persons who reported using insulin alone were classified as having type 1 diabetes. Undiagnosed diabetes was defined as a fasting plasma glucose concentration of 126 mg/dL and was assumed to be type 2. In the fasting subsample, 31 diabetes cases (types 1 and 2) were identified, representing a prevalence of 0.84% (weighted 95% confidence interval (CI): 0.51, 1.40) (276,638 cases; 95% CI: 134,255, 419,020). Estimates of the prevalences of type 1 and type 2 diabetes were 0.48% (95% CI: 0.23, 1.02) and 0.36% (95% CI: 0.20, 0.67), respectively, indicating that T2DM accounted for 43% of all cases. Further, undiagnosed T2DM prevalence was 0.12% (95% CI: 0.05, 0.31), representing 34% of T2DM cases (40,611 cases; 95% CI: 2,850, 78,373). T2DM accounts for approximately half of adolescent diabetes in the United States, and one-third of these cases are undiagnosed. adolescents; diabetes mellitus; etiology; prevalence; undiagnosed diabetes Abbreviations: CDC, Centers for Disease Control and Prevention; CI, confidence interval; MODY, maturity-onset diabetes of the young; NHANES, National Health and Nutrition Examination Survey; SEARCH, SEARCH for Diabetes in Youth Study; T1DM, type 1 diabetes mellitus; T2DM, type 2 diabetes mellitus. Diabetes mellitus is a serious disease which is linked to substantial morbidity and mortality in the United States and globally (1). Type 2 diabetes mellitus (T2DM) among adolescents is particularly concerning, since recent reports suggest that early-onset T2DM is challenging to control and represents a more aggressive disease phenotype than the later-onset form (i.e., more rapid decline in -cell function and accrual of comorbid conditions) (2-4). Therefore, increased prevalence and incidence of early-onset T2DM would portend increased health-care costs, morbidity, and premature mortality (5, 6). Although evidence suggests that the prevalence of T2DM has been increasing among US adolescents (7-16) as a result of the overweight/obesity epidemic (17-19), there are very few reports from population-based studies that can provide reliable estimates, because the overall prevalence of adolescent diabetes remains very low (<1%) (7, 8). In the largest nationally representative sample examining T2DM prevalence in US adolescents to date, Fagot-Campagna et al. (7) reported a total of only 4 cases, 2 of which were undiagnosed, making it difcult to generate precise upper and lower prevalence bounds or to consider variation by sex or race/ethnicity. Alternatively, the SEARCH for Diabetes in Youth Study (SEARCH) has identied more than 6,000 (9) cases of diabetes in youth through an innovative study design, which has attempted to identify all cases of diabetes in dened geographic areas (20). However, the SEARCH sample was recruited 1106 Am J Epidemiol. 2013;178(7):1106-1113 Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 Initially submitted January 18, 2013; accepted for publication April 16, 2013. Diabetes Prevalence Among Adolescents 1107 but the Roche Modular P Chemistry Analyzer (Roche Diagnostics) was used. In all survey cycles and laboratories, fasting glucose was measured according to a hexokinase enzymatic method (24). All glucose values measured from 2005 to 2010 were standardized to the 1999-2004 values using regression equations provided in the analytical notes of the NHANES data documentation for glucose measurement (25, 26). Diabetes classification The Continuous NHANES is a complex, multistage probability sample of US noninstitutionalized civilians that began in 1999. It consists of 6 unique data sets that have been generated in 2-year cycles (i.e., 1999-2000, 2001-2002, 2003- 2004, 2005-5006, 2007-2008, and 2009-2010) (23). Each 2-year survey cycle examines a nationally representative sample of approximately 10,000 persons and collects a variety of health-related data via questionnaire, physical examination, and laboratory assessments. Participation rates for the 6 survey cycles have ranged from 78% to 84%. The current analysis included 11,888 adolescents aged 12- 19 years who responded to a diabetes questionnaire (the interview sample). Among these 11,888 adolescents, a randomly selected subgroup of 4,661 also provided a fasting blood sample (the fasting subsample). Therefore, all 11,888 participants responded to interviewer-administered questionnaires, while only 4,661 participants also provided a fasting blood sample. Responses to interviewer-administered questionnaires were provided by a proxy for participants under 16 years of age; all other participants provided self-report responses. The NHANES protocol was approved by the National Center for Health Statistics institutional review board, and written informed consent was obtained from all participants. Diagnosed diabetes was dened on the basis of a participant self-report of a previous diabetes diagnosis by a physician or health professional, as has been done in previous studies (7, 8). Two respondents who reported \"no\" or \"borderline\" for diabetes but reported that they were taking \"diabetic pills to lower blood sugar\" were conservatively classied as diabetesfree. Participants who reported either no medication use or use of any blood glucose-lowering medication (with or without insulin) were dened as having T2DM, while participants who reported only insulin use were dened as having type 1 diabetes mellitus (T1DM). This approach differs from previously used epidemiologic case denitions of diabetes (7, 8) in that participants receiving both insulin and oral hypoglycemic agents were classied as having T2DM, whereas in previous studies all subjects receiving any insulin therapy had been classied as having T1DM. Approximately 25%-30% of persons with T2DM are treated with insulin (usually along with oral hypoglycemic agents) (27, 28), while very few persons with T1DM are treated with oral hypoglycemic agents (29, 30). Therefore, our current approach, while not perfect, was less likely to underestimate the prevalence of T2DM while having little impact on the underestimation of T1DM, as previously discussed (14). In the fasting subsample, diagnosed diabetes was dened as above (irrespective of fasting glucose values) and undiagnosed diabetes was dened by a fasting blood glucose concentration of 126 mg/dL among participants who did not report a previous diabetes diagnosis during the interview (31). All undiagnosed diabetes cases were classied as T2DM (7). Estimates of the prevalence of diagnosed diabetes were derived from both the interview sample (diagnosed) and the fasting subsample (undiagnosed) for comparative purposes, but we based all nal inferences regarding the prevalence of undiagnosed diabetes on the results obtained from the interview sample because the sample size was approximately 3 times larger, resulting in more precise estimates. Fasting glucose assessments Statistical analysis Fasting status was veried via interview prior to blood collection. During the years 1999-2004, glucose measurements for NHANES were performed by the Diabetes Diagnostic Laboratory at the University of Missouri-Columbia using the Cobas Mira Chemistry System (Roche Diagnostic Systems, Inc., Montclair, New Jersey). In 2005-2006, measurements were performed by the Fairview Medical Center Laboratory at the University of Minnesota using a Roche/ Hitachi 911 Analyzer (Roche Diagnostics, Indianapolis, Indiana). In the NHANES 2007-2010 survey cycle, glucose measurements were still performed at the University of Minnesota, Survey procedures in SAS, version 9.3 (SAS Institute, Inc., Cary, North Carolina), were used for all analyses to obtain the correct variance estimates and corresponding 95% condence intervals. Publicly available survey weights were generated for all 6 continuous NHANES survey cycles by the Centers for Disease Control and Prevention (CDC). These survey weights were used to create a nal analysis survey weight, per the CDC protocol (32), enabling the generation of prevalence estimates and absolute counts representative of US adolescents aged 12-19 years at the midpoint of the combined 12-year survey period. Survey weights are necessary MATERIALS AND METHODS Am J Epidemiol. 2013;178(7):1106-1113 Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 from 6 distinct geographic areas (9) that are not completely representative of the US adolescent population, and estimates of undiagnosed diabetes cannot be obtained from SEARCH. Obtaining nationally representative, reliable prevalence estimates for undiagnosed T2DM is important because it is possible that some of the reported increase in T2DM prevalence might be due to changing diagnostic standards across time and region (diagnostic bias), as well as increased awareness resulting in surveillance bias (16). The literature on autism incidence provides a recent and compelling example of how diagnostic and surveillance bias can misrepresent secular trends in pediatric conditions (21, 22). An accurate prevalence estimate of adolescent diabetes is necessary for informing public health prevention efforts and screening policy. The purpose of this study was to determine the prevalence of diagnosed and undiagnosed T2DM among adolescents. These results arise from 6 nationally representative serial cross-sections of US adolescents enrolled in the Continuous National Health and Nutrition Examination Survey (NHANES) during a 12year time period from 1999 to 2010. 1108 Demmer et al. RESULTS The interview sample included 11,888 adolescents aged 12-19 years who provided self-report information on diabetes status, representing 32,676,471 US adolescents. Weighted for the US population, the participants analyzed were 49% female, 61% white, 12% Mexican-American, 6% other Hispanic, 14% non-Hispanic black, and 7% other race/ethnicity and had a mean age of 15.4 years (standard error, 0.04). Among the subgroup of 4,661 adolescents (representing 32,746,826 US adolescents) who also provided fasting blood samples (the fasting sample), the mean fasting glucose concentration was 92.6 mg/dL (standard error, 0.5). The sex and race/ethnicity distributions in the fasting subsample were similar to those in the full sample (data not shown). Diagnosed diabetes in the interview sample The total number of diagnosed diabetes cases detected in the 6 serial cross-sections of the Continuous NHANES (1999- 2010) was 58, representing 182,960 (weighted 95% condence interval (CI): 119,963, 245,956) diagnosed cases of diabetes. The prevalence was 0.56% (95% CI: 0.4, 0.8) or 5.6 diagnosed diabetes cases per 1,000 US adolescents as of the decade's midpoint. Estimates of diabetes prevalence among males and females were 0.69% versus 0.42% (P = 0.22), and prevalence varied according to race/ethnicity (Table 1), although this nding was not statistically signicant (P = 0.44). The absolute numbers of diagnosed T1DM and T2DM cases detected were 30 and 28 (for the unweighted frequency distribution according to sex and race/ethnicity, see Web Table 1, available at http://aje.oxfordjournals.org/), representing weighted prevalence estimates of 0.38% (95% CI: 0.25, 0.58) and 0.18% (95% CI: 0.10, 0.30), respectively, and indicating that T1DM and T2DM account for 69% and 31% of diagnosed adolescent diabetes. Only 4 persons reported using both insulin and oral medication, and they would have been dened as having T1DM according to previous methods (7, 8), but they were dened as having T2DM here. As is shown in Web Table 2, the inuence of these 4 participants on the overall estimates of self-reported T1DM and T2DM was marginal; when dening the 4 participants using insulin and oral medications as having T1DM, 72% of all diabetes cases were T1DM, while 28% were T2DM. Among these 4 participants, 50% were male, 75% reported their race/ethnicity as non-Hispanic black, and 25% reported their race/ethnicity as Mexican-American. The mean age was 15.25 years (range, 13-17 years), and all 4 participants were obese (i.e., above the 95th percentile of body mass index for age according to CDC clinical growth charts). Total (diagnosed and undiagnosed) diabetes in the fasting subsample Prevalence estimates were higher in the fasting subsample because of undiagnosed diabetes. The overall number of diabetes cases detected was 31, representing a prevalence of 0.84% (95% CI: 0.51, 1.40) or 8.4 adolescents with diabetes per 1,000 US adolescents (for the unweighted frequency distribution by sex and race/ethnicity, see Web Table 3). The weighted prevalence estimates of T1DM and T2DM were 0.48% (95% CI: 0.23, 1.02) and 0.36% (95% CI: 0.2, 0.67), respectively. T1DM and T2DM accounted for 57% and 43% of adolescent diabetes cases, respectively. The seemingly paradoxical nding that the unweighted number of T2DM cases was larger than the unweighted number of T1DM cases, while the reverse was true of the weighted frequencies (Table 2), was due to the higher prevalence of T2DM among participants who were oversampled in various NHANES survey cycles (e.g., among Mexican Americansan oversampled population subgroup in NHANESthe weighted estimated prevalences of T2DM and T1DM were 0.73% and 0%, respectively). Prevalence estimates by sex and race/ethnicity were similar to those in the interview sample, although there was a notable increase in the diabetes prevalence among participants reporting \"other\" race/ethnicity (Table 2). Among adolescents with T2DM, 34% were estimated to be undiagnosed, which represents 40,611 (95% CI: 2,850, 78,373) cases in the general US adolescent population (Table 3). Sex and racial/ethnic variations in the proportion of T2DM that was undiagnosed (Web Figure 1) were not statistically signicant. In comparison with persons with T2DM, those with T1DM had lower body mass indices and were much more likely to have normal weight (Table 4). DISCUSSION We examined 12 years of data from the Continuous NHANES (1999-2010) to provide the rst nationally representative estimates of T2DM, including undiagnosed diabetes, among adolescents in the United States. We found the total prevalence of diabetes (types 1 and 2, diagnosed and undiagnosed) to be 0.84%. The prevalence of T2DM was 0.36%, representing an estimated 119,224 cases of adolescent T2DM in the United States, 34% (>40,000 cases) of which were undiagnosed. T2DM accounted for 43% of total diabetes cases. The current NHANES data suggest a higher prevalence of total diabetes (i.e., 0.84%) in comparison with the estimated Am J Epidemiol. 2013;178(7):1106-1113 Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 to account for nonresponse and oversampling. For example, because NHANES oversamples participants on the basis of characteristics such as race/ethnicity, unweighted prevalence estimates for all US adolescents aged 12-19 years would be biased as a result of the overrepresentation of population subgroups. Statistical hypothesis tests were based on 2-sample t tests for continuous variables and on 2 tests for proportions. The Wilson method was used to estimate 95% condence intervals, as the prevalence estimates were close to zero. Prevalence estimates were also provided for subgroups based on sex and race/ ethnicity. Race/ethnicity was categorized as non-Hispanic white, non-Hispanic black, Mexican-American, other Hispanic, and other (including Asian/Pacic Islander, American Indian, and multiracial). Because Hispanics other than Mexican Americans were not oversampled until 2007-2008, the sample size in this group was not sufcient to provide reliable prevalence estimates; therefore, any estimates presented here should be interpreted cautiously. We opted to present estimates for this subgroup because no data on this demographic group exist to date. The prediction of undiagnosed diabetes by sex and race/ethnicity was explored using logistic regression analyses. Diabetes Prevalence Among Adolescents 1109 Table 1. Prevalence of Self-Reported Diabetes Mellitusa Among 11,888 US Adolescents Aged 12-19 Years, National Health and Nutrition Examination Survey, 1999-2010 Total No. of Personsb Type 1 Diabetesc All Diabetes Type 2 Diabetesc No. of Personsb %d 95% CId No. of Personsb %d 32,676,471 182,960 0.56 0.40, 0.79 125,422 0.38 Male 16,776,646 115,851 0.69 0.44, 1.09 81,460 0.49 0.29, 0.82 34,391 0.21 0.10, 0.43 Female 15,899,824 67,109 0.42 0.23, 0.76 43,962 0.28 0.12, 0.63 23,147 0.15 0.08, 0.28 Non-Hispanic white 19,904,320 113,784 0.57 0.35, 0.93 86,342 0.43 0.26, 0.73 27,442 0.14 0.05, 0.35 Non-Hispanic black 4,758,796 33,470 0.70 0.44, 1.12 13,343 0.28 0.14, 0.57 20,127 0.42 0.23, 0.78 MexicanAmerican 3,757,873 13,807 0.37 0.20, 0.67 5,075 0.14 0.04, 0.44 8,732 0.23 0.12, 0.44 Other Hispanic 2,040,119 18,707 0.92 0.22, 3.61 18,707 0.92 0.23, 3.61 0 Othere 2,215,363 3,192 0.14 0.02, 0.98 1,955 0.08 0.009, 0.88 1,237 0.06 0.004, 0.82 Total population 95% CId 0.25, 0.58 No. of Personsb %d 57,538 0.18 95% CId 0.10, 0.30 Sex Race/ethnicity prevalence of 0.41% in the Third National Health and Nutrition Examination Survey (NHANES III; 1988-1994) (7), and the proportion of total diabetes that is type 2 also appears to be higher (31% in NHANES III vs. 43% presently). However, comparisons with NHANES III estimates should be interpreted very cautiously, because the NHANES III ndings were based on a total of only 13 diabetes cases, which resulted in low precision for the total diabetes prevalence estimate (reected by a 95% condence interval that included 0) (7). Moreover, the total number of T2DM cases detected in NHANES III was only 4 (2 undiagnosed), precluding the investigators' ability to provide a precise prevalence estimate of diagnosed or undiagnosed T2DM or to explore variation by sex or race/ethnicity (7). Therefore, it was not possible for us to compare the current NHANES data with data from NHANES III and determine whether the increased prevalence was driven by chance versus an actual increase in T2DM, T1DM, or both. When only diagnosed diabetes was considered, we found the prevalences of total diagnosed diabetes (type 1 or type 2) and diagnosed T2DM to be 0.56% and 0.18%, respectively. These estimates were based on the interview sample as opposed to the fasting subsample, because the sample size was nearly 3-fold greater in the interview sample, which enhanced the precision of prevalence estimates. In comparison with a previous single NHANES cross-section from 1999-2002, Duncan Am J Epidemiol. 2013;178(7):1106-1113 (8) estimated the prevalence of total diagnosed diabetes (type 1 or type 2) to be 0.5% and the prevalence of diagnosed T2DM to be 0.15%gures which are very comparable with, yet still lower than, our current estimates. The current estimates are also higher than those previously published by the SEARCH investigators (approximately 0.3%) in a similar age group (9). The increased prevalence of T2DM currently reported in NHANES 1999-2010 (12 years) in comparison with NHANES 1999-2002 or SEARCH might reect a true increase in prevalence, which would be consistent with trends of increasing obesity and body mass index, albeit very modest ones, occurring from 1999-2000 through 2009-2010 (18). It is also possible that the higher prevalence of diagnosed T2DM observed presently reects increased awareness and screening among clinicians, which has in turn decreased the proportion of T2DM that goes undiagnosed. Unfortunately, the very low number of undiagnosed T2DM cases observed in any specic NHANES cross-section precludes reliable estimation of secular trends in these data. A comparison of undiagnosed diabetes prevalence between NHANES 1999-2010 and SEARCH cannot be made, because SEARCH does not identify undiagnosed disease. Alternatively, while the SEARCH study is remarkable in its scope, it is not representative of all US geographic areas (9), since it would be untenable to identify all diagnosed diabetes cases in the United States. Therefore, the lower prevalence observed in SEARCH might reect a truly lower Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 Abbreviation: CI, confidence interval. a Respondents (adolescents aged 12-19 years) were asked, \"Have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?\" Adolescents who responded \"yes\" and reported using only insulin were defined as having type 1 diabetes; adolescents who responded \"yes\" and reported either no medication use or use of oral diabetic pills (with or without insulin) to lower blood sugar were defined as having type 2 diabetes. b Number of noninstitutionalized US adolescents weighted for the US population. Differences from totals are due to rounding. c Unweighted frequencies of observed diabetes cases by etiology in the sample were as follows: type 1 diabetes, n = 30; type 2 diabetes, n = 28. d Percentage of noninstitutionalized US adolescents (and 95% confidence interval) weighted for the US population. e Including Asian/Pacific Islander, American Indian, and multiracial. 1110 Demmer et al. Table 2. Total Prevalence of All Diabetes Mellitus (Diagnosed and Undiagnosed) Among 4,661 US Adolescents Aged 12-19 Years, National Health and Nutrition Examination Survey, 1999-2010 Type 1 Diabetesb All Diabetes Total No. of Personsa No. of Personsa %c No. of Personsa %c 32,746,826 276,638 0.84 0.51, 1.40 157,413 0.48 Male 16,790,805 160,599 Female 15,956,021 116,039 0.96 0.51, 1.77 86,076 0.73 0.30, 1.73 71,337 Non-Hispanic white 20,201,342 145,421 0.72 0.36, 1.42 Non-Hispanic black 4,760,434 35,241 0.74 MexicanAmerican 3,732,743 27,386 Other Hispanic 1,880,888 Otherd 2,171,419 Total population Type 2 Diabetesb No. of Personsa %c 95% CIc 0.23, 1.02 119,224 0.36 0.20, 0.67 0.51 0.21, 1.22 74,523 0.44 0.19, 1.02 0.45 0.12, 1.60 44,701 0.28 0.13, 0.60 89,249 0.44 0.18, 1.06 56,170 0.28 0.10, 0.78 0.34, 1.61 16,338 0.34 0.11, 1.05 18,904 0.40 0.14, 1.14 0.73 0.39, 1.38 0 27,386 0.73 0.39, 1.38 51,826 2.75 0.60, 11.67 16,764 0.77 0.17, 3.40 0.77 0.17, 3.60 95% CIc 95% CIc Sex Race/ethnicity 2.80 0.60, 11.67 0 0 16,764 Abbreviation: CI, confidence interval. a Number of noninstitutionalized US adolescents weighted for the US population. Differences from totals are due to rounding. b Unweighted frequencies of observed diabetes cases by etiology in the sample were as follows: type 1 diabetes, n = 10; type 2 diabetes, n = 21. c Percentage of noninstitutionalized US adolescents (and 95% confidence interval) weighted for the US population. d Including Asian/Pacific Islander, American Indian, and multiracial. prevalence in the geographical regions covered by the study. It is also possible that the self-reported nature of diagnosed diabetes in NHANES leads to overreporting by adolescents (or their proxies). For example, youths who were told by a health-care provider that they either had prediabetes or were at risk of developing future diabetes might have misunderstood that they actually had diabetes. In regard to our estimate of undiagnosed T2DM, some important limitations should be noted. It is possible that our estimates were biased because we had only 1 fasting glucose Table 3. Proportions of Diagnosed and Undiagnosed Type 2 Diabetes Mellitus Among US Adolescents Aged 12-19 Years, by Sex and Race/ Ethnicity, National Health and Nutrition Examination Survey, 1999-2010 All T2DMa Diagnosed T2DMa Undiagnosed T2DMa No. of Personsb %c 95% CIc No. of Personsb %c 95% CIc No. of Personsb %c 95% CIc 119,224 0.36 0.20, 0.67 78,613 0.24 0.11, 0.51 40,611 0.12 0.05, 0.31 Male 74,523 0.44 0.19, 1.02 44,087 0.26 0.08, 0.82 30,436 0.18 0.06, 0.55 Female 44,701 0.28 0.13, 0.60 34,526 0.22 0.09, 0.53 10,175 0.06 0.01, 0.29 Non-Hispanic white 56,171 0.28 0.10, 0.80 44,783 0.22 0.07, 0.72 11,388 0.06 0.01, 0.40 Non-Hispanic black 18,904 0.40 0.14, 1.14 12,819 0.27 0.07, 1.07 6,084 0.13 0.03, 0.50 Mexican-American 27,386 0.73 0.40, 1.40 17,737 0.48 0.22, 1.04 9,649 0.26 0.09, 0.75 0.62 0.11, 3.45 Total population Sex Race/ethnicity Other Hispanic Otherd 0 16,764 0 0.77 0.17, 3.40 3,274 0 0.15 0.01, 2.04 13,490 Abbreviations: CI, confidence interval; T2DM, type 2 diabetes mellitus. a Unweighted frequencies of observed type 2 diabetes cases in the sample were as follows: total diabetes, n = 21; diagnosed type 2 diabetes, n = 13; undiagnosed type 2 diabetes, n = 8. b Number of noninstitutionalized US adolescents weighted for the US population. Differences from totals are due to rounding. c Percentage of noninstitutionalized US adolescents (and 95% confidence interval) weighted for the US population. d Including Asian/Pacific Islander, American Indian, and multiracial. Am J Epidemiol. 2013;178(7):1106-1113 Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 51,826 Diabetes Prevalence Among Adolescents 1111 Table 4. Characteristicsa of 4,661 Fasting US Adolescents Aged 12-19 Years According to Diabetes Mellitus Status and Type, National Health and Nutrition Examination Survey, 1999-2010 No Diabetesb Mean (SE) Range Age, years 15.5 (0.05) BMIc 23.4 (0.1) BMI z score T1DMb % Mean (SE) Range 12-19 16.8 (0.7) 13-57 22.7 (1.0) 0.6 (0.03) 3.9 to 3.3 Diagnosed T2DMb % Mean (SE) Range 12-19 16.0 (0.6) 18-26 29.1 (3.3) 0.5 (0.2) 1.4 to 1.2 1.3 (0.5) Undiagnosed T2DMb % Mean (SE) Range 13-19 15.8 (0.8) 14-19 21-46 35.7 (3.9) 20-47 0.1 to 2.9 2.1 (0.4) 0.7 to 3.0 % BMI category Overweightd 16 19 25 0 Obesee 18 0 36 79 Fasting plasma glucose, mg/dL 91 (0.2) 58-123 314 (46) 65-517 155 (27) 81-362 182 (14) 128-249 value and no glucose challenge data, which have enhanced sensitivity for diagnosing T2DM in children (33). While our approach is commonly used in epidemiologic studies, it does not meet the specic American Diabetes Association criteria, which require repeat testing in the absence of unequivocal hyperglycemia (31). Utilization of the NHANES hemoglobin A1c data might have modestly improved our sensitivity, but unfortunately there was an unexplained shift of hemoglobin A1c distributions in 2007-2010 versus 1999-2006 reported in the online NHANES documentation (34), which raised as many problems as solutions for our current aims. Nevertheless, recent studies have demonstrated that fasting glucose measurement has enhanced T2DM diagnostic accuracy among youth when compared with hemoglobin A1c level (33). Overestimation of undiagnosed disease is also possible if some participants did not fast, although fasting status was veried via questionnaire for all participants prior to blood collection, which minimized the potential for this bias. There appear to be possible disparities in the prevalences of both diagnosed and undiagnosed diabetes by sex, with males having a higher prevalence of diabetes overall as well as a higher prevalence of diagnosed and undiagnosed T2DM. The results also suggested possible racial/ethnic disparities that would be consistent with diabetes trends among adults (35). However, neither the sex difference nor the racial/ethnic differences were statistically signicant, which makes it possible that these variations were due to chance. Studies with higher numbers of diabetes cases will be required to denitively examine these trends. Presently, self-reported medication use was incorporated into our denition of diabetes type. In situations where participants report use of both insulin and oral hypoglycemic agents, the best method for dening T1DM versus T2DM is unclear. We chose to dene persons who reported insulin use combined with oral medication use as having type 2, whereas Am J Epidemiol. 2013;178(7):1106-1113 previous reports dened them as type 1. We believe our approach is less likely to misclassify T2DM cases as T1DM (28), with minimal risk of misclassifying T1DM cases as T2DM. Therefore, our denition probably enhances sensitivity without sacricing specicity, and our supplemental analyses alternatively dening anyone using both medications as type 1 did not meaningfully change the results in the full sample. However, there was evidence to suggest that the method for dening diabetes type could be important among certain racial/ ethnic groups, as well as among obese adolescents. Among the 4 participants who self-reported using both insulin and oral medication, 3 were non-Hispanic black, one was Mexican-American, and all 4 were obese. Future research that can inform the prevalence of combined insulin and oral medication use among persons whose diabetes type has been rigorously validated will be important to help ensure optimal and standardized denitions over time and across research settings. While most T2DM among adolescents is related to overweight/obesity, we opted not to utilize this information in dening etiological type, because 33% of SEARCH participants with T1DM also have body mass index values at the 85th percentile or above (36). Therefore, using adiposity as an etiological criterion in an epidemiologic setting where overweight/obesity is highly prevalent would probably misclassify many cases of T1DM as T2DM and would therefore overestimate T2DM prevalence. The lack of detailed information on diabetes etiology in NHANES also precluded the identication of maturity-onset diabetes of the young (MODY) in these data, although the inuence of misclassied MODY on the current prevalence estimates is likely to have been minimal, since MODY accounts for less than 5%-10% of youth-onset diabetes (37). Despite limitations regarding our diabetes case denition and etiology classication in addition to the inability to present reliable secular trends, we believe these data are of Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 Abbreviations: BMI, body mass index; SE, standard error; T1DM, type 1 diabetes mellitus; T2DM, type 2 diabetes mellitus. a Weights from the National Health and Nutrition Examination Survey were used to obtain results representative of US adolescents. b Unweighted frequencies of observed diabetes cases in the sample, overall and by etiology, were as follows: no diabetes, n = 4,630; type 1 diabetes, n = 10; diagnosed type 2 diabetes, n = 13; undiagnosed type 2 diabetes, n = 8. c Weight (kg)/height (m)2. d Overweight (85th-94th percentile of BMI for age) was defined according to Centers for Disease Control and Prevention clinical growth charts. e Obesity (95th percentile of BMI for age) was defined according to Centers for Disease Control and Prevention clinical growth charts. 1112 Demmer et al. ACKNOWLEDGMENTS Author afliations: Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York (Ryan T. Demmer, Aleksandra M. Zuk, Mose Desvarieux); Division of Molecular Genetics, Departments of Pediatrics and Medicine, College of Physicians and Surgeons, Columbia University, New York, New York (Michael Rosenbaum); INSERM, U738, Paris, France (Mose Desvarieux); and School of Advanced Studies in Public Health, Paris, France (Mose Desvarieux). This research was supported by National Institutes of Health grant R00 DE-018739 (Dr. Demmer); an award (DK 63608) to Dr. Demmer from the Diabetes and Endocrinology Research Center, College of Physicians and Surgeons, Columbia University; and National Institutes of Health grant R01 DE-13094 (Dr. Desvarieux). Dr. Desvarieux also received support from a chair in chronic disease at the cole des Hautes tudes en Sant Publique, Paris, France. The funders played no role in the study design, data collection, and analysis, the decision to publish, or the preparation of the manuscript. Conict of interest: none declared. REFERENCES 1. Wild S, Roglic G, Green A, et al. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care. 2004;27(5):1047-1053. 2. Zeitler P, Hirst K, Pyle L, et al. A clinical trial to maintain glycemic control in youth with type 2 diabetes. N Engl J Med. 2012;366(24):2247-2256. 3. Song SH, Hardisty CA. Early onset type 2 diabetes mellitus: a harbinger for complications in later yearsclinical observation from a secondary care cohort. QJM. 2009;102(11):799-806. 4. Song SH, Hardisty CA. Early-onset type 2 diabetes mellitus: an increasing phenomenon of elevated cardiovascular risk. Expert Rev Cardiovasc Ther. 2008;6(3):315-322. 5. Narayan KM, Boyle JP, Thompson TJ, et al. Lifetime risk for diabetes mellitus in the United States. JAMA. 2003;290(14): 1884-1890. 6. Daneman D. Early diabetes-related complications in adolescents: risk factors and screening. Horm Res. 2005; 63(2):75-85. 7. Fagot-Campagna A, Saaddine JB, Flegal KM, et al. Diabetes, impaired fasting glucose, and elevated HbA1c in U.S. adolescents: the Third National Health and Nutrition Examination Survey. Diabetes Care. 2001;24(5):834-837. 8. Duncan GE. Prevalence of diabetes and impaired fasting glucose levels among US adolescents: National Health and Nutrition Examination Survey, 1999-2002. Arch Pediatr Adolesc Med. 2006;160(5):523-528. 9. Liese AD, D'Agostino RB Jr, Hamman RF, et al. The burden of diabetes mellitus among US youth: prevalence estimates from the SEARCH for Diabetes in Youth Study. Pediatrics. 2006;118(4):1510-1518. 10. Fagot-Campagna A. Emergence of type 2 diabetes mellitus in children: epidemiological evidence. J Pediatr Endocrinol Metab. 2000;13(suppl 6):1395-1402. 11. Weill J, Vanderbecken S, Froguel P. Understanding the rising incidence of type 2 diabetes in adolescence. Arch Dis Child. 2004;89(6):502-504. 12. Mohamadi A, Cooke DW. Type 2 diabetes mellitus in children and adolescents. Adolesc Med State Art Rev. 2010;21(1):103-119. 13. Liese AD, Lawson A, Song HR, et al. Evaluating geographic variation in type 1 and type 2 diabetes mellitus incidence in youth in four US regions. Health Place. 2010;16(3):547-556. 14. Nadeau K, Dabelea D. Epidemiology of type 2 diabetes in children and adolescents. Endocr Res. 2008;33(1-2):35-58. 15. Gungor N, Hannon T, Libman I, et al. Type 2 diabetes mellitus in youth: the complete picture to date. Pediatr Clin North Am. 2005;52(6):1579-1609. 16. Eko JM. The Epidemiology of Diabetes Mellitus. 2nd ed. Chichester, United Kingdom: Wiley-Blackwell; 2008. 17. Ogden CL, Flegal KM, Carroll MD, et al. Prevalence and trends in overweight among US children and adolescents, 1999-2000. JAMA. 2002;288(14):1728-1732. 18. Ogden CL, Carroll MD, Kit BK, et al. Prevalence of obesity and trends in body mass index among US children and adolescents, 1999-2010. JAMA. 2012;307(5):483-490. 19. Rosenbaum M. Epidemiology of pediatric obesity. Pediatr Ann. 2007;36(2):89-95. 20. SEARCH Study Group. SEARCH for Diabetes in Youth: a multicenter study of the prevalence, incidence and classication of diabetes mellitus in youth. Control Clin Trials. 2004;25(5):458-471. 21. Barbaresi WJ, Colligan RC, Weaver AL, et al. The incidence of clinically diagnosed versus research-identied autism in Olmsted County, Minnesota, 1976-1997: results from a retrospective, population-based study. J Autism Dev Disord. 2009;39(3):464-470. 22. Barbaresi WJ, Katusic SK, Colligan RC, et al. The incidence of autism in Olmsted County, Minnesota, 1976-1997: results from a population-based study. Arch Pediatr Adolesc Med. 2005;159(1):37-44. 23. National Center for Health Statistics, Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey. Questionnaires, Datasets, and Related Documentation. Hyattsville, MD: National Center for Health Statistics; 2012. (http://www.cdc.gov/nchs/nhanes/ nhanes_questionnaires.htm). (Accessed April 5, 2013). 24. National Center for Health Statistics, Centers for Disease Control and Prevention. Laboratory Procedure Manual. Am J Epidemiol. 2013;178(7):1106-1113 Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 high value given the fact that no other study can provide nationally representative benchmark estimates of total (diagnosed and undiagnosed) T2DM among adolescents. We found the prevalence of T2DM during the period 1999-2010 to be 0.36% in a nationally representative sample of US adolescents, and 34% of T2DM appeared to be undiagnosed. The overall prevalence of adolescent diabetes was 0.84%, which suggests an increase in comparison with the prevalence observed in the late 1980s/early 1990s. The degree to which the apparent increase in both diagnosed and overall diabetes prevalence is attributable to diagnostic bias, surveillance bias, and/or true increases in adolescent T2DM incidence remains to be determined in future studies. Nevertheless, the data reported presently from NHANES 1999-2010 serve as an important benchmark against which future studies can be compared to better inform the burden of diabetes among adolescents, particularly as the joint pandemics of obesity and diabetes continue to evolve in the United States and globally. Diabetes Prevalence Among Adolescents 1113 25. 26. 28. 29. 30. Am J Epidemiol. 2013;178(7):1106-1113 31. American Diabetes Association. Diagnosis and classication of diabetes mellitus. Diabetes Care. 2012;35(suppl 1): S64-S71. 32. National Center for Health Statistics, Centers for Disease Control and Prevention. Specifying Weight Parameters. Hyattsville, MD: National Center for Health Statistics; 2011. (http://www.cdc.gov/nchs/tutorials/NHANES/SurveyDesign/ Weighting/intro.htm). (Accessed April 10, 2013). 33. Nowicka P, Santoro N, Liu H, et al. Utility of hemoglobin A1c for diagnosing prediabetes and diabetes in obese children and adolescents. Diabetes Care. 2011;34(6):1306-1311. 34. National Center for Health Statistics, Centers for Disease Control and Prevention. Analytic notes. In: National Health and Nutrition Examination Survey. 2007-2008 Data Documentation, Codebook, and Frequencies. Glycohemoglobin (GHB_E). Hyattsville, MD: National Center for Health Statistics; 2012. (http://www.cdc.gov/nchs/nhanes/ nhanes2007-2008/GHB_E.htm#Analytic_Notes). (Accessed July 9, 2013). 35. Cowie CC, Rust KF, Byrd-Holt DD, et al. Prevalence of diabetes and impaired fasting glucose in adults in the U.S. population: National Health and Nutrition Examination Survey 1999-2002. Diabetes Care. 2006;29(6): 1263-1268. 36. Dabelea D, Pihoker C, Talton JW, et al. Etiological approach to characterization of diabetes type: the SEARCH for Diabetes in Youth Study. Diabetes Care. 2011;34(7):1628-1633. 37. Winter WE, Nakamura M, House DV. Monogenic diabetes mellitus in youth. The MODY syndromes. Endocrinol Metab Clin North Am. 1999;28(4):765-785. Downloaded from http://aje.oxfordjournals.org/ at Harvard University on March 16, 2014 27. Hyattsville, MD: National Center for Health Statistics; 2009. (http://www.cdc.gov/nchs/data/nhanes/nhanes_07_08/ glu_e_met_fasting_glucose_ModP.pdf ). (Accessed December 10, 2011). National Center for Health Statistics, Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey. 2005-2006 Data Documentation, Codebook, and Frequencies. Plasma Fasting Glucose and Insulin (GLU_D). Hyattsville, MD: National Center for Health Statistics; 2008. (http://www.cdc.gov/nchs/nhanes/ nhanes2005-2006/GLU_D.htm). (Accessed April 5, 2013). National Center for Health Statistics, Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey. 2007-2008 Data Documentation, Codebook, and Frequencies. Plasma Fasting Glucose and Insulin (GLU_E). Hyattsville, MD: National Center for Health Statistics; 2010. (http://www.cdc.gov/nchs/nhanes/ nhanes2007-2008/GLU_E.htm). (Accessed April 5, 2013). Koro CE, Bowlin SJ, Bourgeois N, et al. Glycemic control from 1988 to 2000 among U.S. adults diagnosed with type 2 diabetes: a preliminary report. Diabetes Care. 2004;27(1): 17-20. Elder DA, Herbers PM, Weis T, et al. Beta-cell dysfunction in adolescents and adults with newly diagnosed type 2 diabetes mellitus. J Pediatr. 2012;160(6):904-910. Adi S. Type 1 diabetes mellitus in adolescents. Adolesc Med State Art Rev. 2010;21(1):86-102, ix. Jovanovic L. Point: Oral hypoglycemic agents should not be used to treat diabetic pregnant women. Diabetes Care. 2007;30(11):2976-2979. 1. What is the research question (aim of the research)? 2. Describe how the participants for this study were chosen. 3. What type of epidemiologic study design is this? 4. Was this an appropriate design give the aim of the study? 5. What was the main outcome (dependent variable) of interest and what type of variable is it? How was the outcome assessed/operationally defined for the analysis? 6. What were the independent variables included in the main analyses? How were they operationally defined/assessed? What types of variables are they? 7. What are the key findings from table 1? 8. How does table 3 differ from table 1? 9. What are the key points from table 3? 10. What are the potential limitations of this study