UC Davis

Objective

To test the hypothesis that clinical observations made at patient presentation can distinguish type 2 diabetes (T2D) from type 1 diabetes (T1D) in pediatric patients aged 2 to 18.

Subjects and methods

Medical records of 227 African American and 112 Hispanic American pediatric patients diagnosed as T1D or T2D were examined to compare parameters in the two diseases. Age at presentation, BMI z-score, and gender were the variables used in logistic regression analysis to create models for T2D prediction.

Results

The regression-based model created from African American data had a sensitivity of 92% and a specificity of 89%; testing of a replication cohort showed 91% sensitivity and 93% specificity. A model based on the Hispanic American data showed 92% sensitivity and 90% specificity. Similarities between African American and Hispanic American patients include: (1) age at onset for both T1D and T2D decreased from the 1980s to the 2000s; (2) risk of T2D increased markedly with obesity. Racial/ethnic-specific observations included: (1) in African American patients, the proportion of females was significantly higher than that of males for T2D compared to T1D (p<0.0001); (2) in Hispanic Americans, the level of glycated hemoglobin (HbA1c) was significantly higher in T1D than in T2D (p<0.002) at presentation; (3) the strongest contributor to T2D risk was female gender in African Americans, while the strongest contributor to T2D risk was BMI z-score in Hispanic Americans.

Conclusions

Distinction of T2D from T1D at patient presentation was possible with good sensitivity and specificity using only three easily-assessed variables: age, gender, and BMI z-score. In African American pediatric diabetes patients, gender was the strongest predictor of T2D, while in Hispanic patients, BMI z-score was the strongest predictor. This suggests that race/ethnic specific models may be useful to optimize distinction of T1D from T2D at presentation.