Thirty years ago, T2D was rare in children. Generally, children presenting with elevated blood glucose were considered T1D patients and adults presenting with elevated blood glucose were considered T2D patients. In the interim, previously unrecognized forms of diabetes, including MODY (maturity onset of diabetes in youth) and LADA (latent autoimmune diabetes in adults), have blurred the differential diagnosis of diabetes to some extent. The biggest contributor to difficulty in the differential diagnosis of T1D vs. T2D, however, is decreasing age of onset for both T1D and T2D, but particularly for T2D. As a consequence of the decreasing age at presentation over the last two decades, the age at presentation of T2D overlapped that of T1D in both African American and Hispanic American patients. Decreasing age of T1D onset in our data is similar to that of several long-term studies in England, Finland, and Italy 
. A dramatic decrease in the age of onset for T2D has been noted in several studies and is particularly striking in non-Caucasian groups, such as African Americans, Hispanics, and Native Americans, who are, for reasons not yet fully understood, particularly susceptible to T2D 
. This makes distinction of T1D vs. T2D particularly challenging in these groups and is the basis for the current study.
Overweight/obesity was a strong predictor of T2D in both African American and Hispanic American youth; however, the odds ratio for BMI z-score was far greater in HA (25.6) than in AA (4.1). Many lines of evidence support the notion that race/ethnicity can be an important factor in T2D susceptibility. Prevalence of obesity varies among ethnic groups. In a recent study, prevalence of obesity among 4 year old U.S. children was reported as: 31% in American Indian/Native Alaskan, 22% in Hispanic American, 21% in African American, 16% in European American, and 13% in Asian American 
. Among youth with T2D, Asians had a lower mean BMI than did Pacific Islanders, 33.7 vs. 42.4, suggesting that Asian youth may be at risk for T2D at a lower BMI than other racial/ethnic groups 
. Similarly, a study of adults found that South Asian T2D patients had lower BMI than did European white T2D patients 
. Thus, although BMI is strongly correlated with T2D, other factors, perhaps genetic or cultural, combine with BMI to determine susceptibility. Our data show that the predictive test we are developing performs much better when used for patients who belong to the group with which it was created, i.e., the predictive value for HA patients is better when the HA-specific model is used, rather than the AA-specific model. Taken together, all of these observations underscore the fact that race/ethnicity plays a role in T2D susceptibility and should be taken into account when attempting to distinguish diabetes type.
The uneven gender distribution in African American T2D reported here, with significantly more females than males, is consistent with other studies 
. The absence of this gender bias in Hispanic American patients is also consistent with previous reports 
. The difference in gender distribution between T1D and T2D in African American but not in Hispanic American patients, the lesser prominence of BMI in risk of T2D in African American patients, and the difference in T1D and T2D glycated hemoglobin in Hispanic American patients at onset may all reflect differences in underlying genetic influences in the two populations in this study.
Other research supports the notion of race-specific genetic differences that could affect T2D risk. Examination of healthy pediatric subjects showed higher insulin secretion and lower insulin clearance in African American subjects compared to European American subjects 
. The metabolic syndrome, which can precede type 2 diabetes, has different characteristics in European Americans, African Americans, and Hispanic Americans 
. Fatty acid metabolism is different in African Americans compared to European Americans due to a genetic difference in a key enzyme 
This study demonstrates that only a few variables recorded at presentation may provide a model that could aid in differential diagnosis of T1D and T2D in African American and Hispanic American pediatric patients at the time of presentation. This study also provides evidence that models to predict T2D from data recorded at presentation should be race/ethnicity specific because the relative risk of T2D as indicated by gender and BMI z-score is different in African American and Hispanic American patients. These results suggest that other racial/ethnic groups may have their own levels of risk for T2D due to age, gender, and BMI z-score.
A larger study, with data collected prospectively, is currently in the planning stage. Data from this survey-type study will be utilized to replicate the present work, improve the accuracy of prediction of T2D, and test whether the addition of more observational variables, such as presence or absence of acanthosis nigricans, can further improve the prediction of diabetes type. The data presented here suggest that an immediate, simple, rapid, and inexpensive test can be developed to predict diabetes type at presentation in underserved minority children and youth.