The report by Demmer et al. (1) estimated the prevalence of diagnosed and undiagnosed diabetes in US adolescents. As investigators for the SEARCH for Diabetes in Youth Study (herein, the SEARCH Study), a US population-based registry of diabetes in youth (2), we acknowledge that providing valid estimates of the burden of diabetes in children, overall and by type, is an important and complex endeavor, but we wish to raise several concerns about the validity and representativeness of these findings.
The authors reported a prevalence of self-reported type 1 diabetes mellitus (T1DM) that is approximately 65% higher than that reported by the SEARCH Study in 2001 (3.8/1,000 vs. 2.3/1,000, respectively) (3), which may be explained by temporal trends in T1DM prevalence, but also by unvalidated overreporting, both of which the authors mentioned. Surprisingly, the prevalence of self-reported type 2 diabetes mellitus (T2DM) (1.8/1,000 among those aged 12–19 years) was 4.3 times higher than previously reported by the SEARCH Study (0.42/1,000 for those aged 10–19 years) and much closer to SEARCH Study estimates for American Indian youth (1.7/1,000), the racial group with the highest prevalence of T2DM.
We believe that the small number of adolescents screened, the imprecise case definition, and misclassification of diabetes type may underlie this. The authors defined T2DM on the basis of self-reported diet therapy or use of oral hypoglycemic agents with or without insulin. We agree that true diabetes cases who do not take insulin are likely to have T2DM; however, up to 11% of 8,879 youth with T1DM in the SEARCH Study are taking oral hypoglycemic agents in addition to insulin (4–6). Because cases of T1DM greatly outnumber those of T2DM, even low rates of misclassification of T1DM as T2DM will result in an overestimate of T2DM prevalence. Because insulin use varies among minority groups (4–6), differential misclassification by racial/ethnic group will likely occur.
Even higher estimates were reported in the fasting subsample, all of which were considered to have undiagnosed T2DM. The authors acknowledged that fasting glucose levels are variable in youth, and that a 1-time elevated value would not result in a diagnosis of diabetes; however, the impact of such problems is less well recognized. In the HEALTHY Study (7), only 1 of the original 6 youth with elevated fasting glucose was confirmed for a true prevalence of 1/6,358 = 0.02%. This compares to an unweighted prevalence of elevated fasting glucose of 0.17% in the National Health and Nutrition Examination Survey (NHANES) (1). In the Princeton Study of 2,501 elementary students (8), 7 youth with “near diabetes” were identified (0.28%), but only 1 was confirmed to have T2DM, for a prevalence of 0.04%. Thus, the prevalence of previously known (n = 2) and undiagnosed (n = 1) T2DM was 0.12%; the weighted prevalence reported by Demmer et al. (1) was 3 times higher (0.36%).
The authors make the case that the NHANES population is representative; however, their sample contains no American Indians or Asian/Pacific Islanders, 2 groups with the highest risk of T2DM, so it is unclear how representative the sample is for US youth at risk for diabetes. The explanation provided by the authors for the lower T2DM prevalence in the SEARCH Study (3) versus NHANES is that the SEARCH Study is not a geographically representative sample. The geographical areas included in the SEARCH Study were not based on random sampling; however, the SEARCH Study population is quite representative of the racial/ethnic, sex, education, and income distributions in the United States (9). Importantly, the SEARCH Study prevalence estimates are based on more than 6,000 cases with diabetes identified from a denominator of approximately 3.5 million youth compared with only 58 cases in fewer than 12,000 youth from NHANES.
The authors conclude that “T2DM accounts for approximately half of adolescent diabetes in the United States …” (1, p.1). The SEARCH Study recently estimated that 18,919 youth aged 10–19 years in the United States had diagnosed T2DM, and 132,315 youth less than 20 years of age had T1DM, and that T2DM accounts for 12.5% of all diabetes cases in youth (9), which is substantially less than estimated by Demmer et al. (1). Given the small number of cases and the numerous problems with case definitions discussed above, we believe that the estimates provided by Demmer et al. based on NHANES data do not accurately reflect the prevalence of diabetes, especially T2DM, in US adolescents.
Conflict of interest: none declared.