The American Diabetes Association (ADA) has recommended screening for type 2 diabetes by fasting plasma glucose (FPG) in children who are overweight (BMI >85th percentile) who have two of the following risk factors: at-risk ethnic minority origin, family history of diabetes in a first- or second-degree relative, or insulin resistance (acanthosis nigricans, polycystic ovarian syndrome, hypertension, or dyslipidemia) (1).
The case for refining the criteria for screening has been made previously (2). In that study, the sensitivity of the criteria was 24%, with a positive predictive value of 3%, i.e., 40 children needed to be tested to yield one abnormal result. In a response to this, Rosenbloom (3) highlighted the need to test the ADA criteria in high-risk populations to establish the strength and risk level of different factors that are influential in the development of type 2 diabetes.
Our institution covers a population where type 2 diabetes in childhood has emerged (4). We describe our experience of applying the ADA criteria and propose a clinical assessment tool to refine the selection of children for screening by the oral glucose tolerance test (OGTT).
In the last 4 years, 66 children have had OGTTs for suspected glucose intolerance. The characteristics of this population were mean age of 12.7 years (range 4.8–17.3), mean BMI standard deviation score 3.0 (0.0–4.6), 71% female, 83% ethnic minority origin (of whom 73% were South Asian and 9% African Caribbean), 88% had acanthosis nigricans, and 67% had a first- or second-degree family history of diabetes. Of these, 13 children had abnormal glucose tolerance (4 diabetes, 8 impaired glucose tolerance [IGT], and 1 impaired fasting glycemia).
Applying the ADA criteria, 11 of the 13 children with abnormal results would have qualified for screening, missing 1 child with diabetes and 1 with IGT. Screening these 11 with FPG as per the recommendations would have missed a further 7 children, 1 with diabetes and 6 with IGT, as only 1 of the children with IGT had impaired fasting glycemia. Overall, the sensitivity of the ADA criteria using FPG as a screening test in our population was 31%. Use of the ADA criteria to screen by OGTT would have identified 11 of the 13 abnormal results in this cohort, giving a sensitivity of 85% for the criteria and a specificity of 26%, with a positive predictive value of 22%, i.e., five children would need to be tested to yield an abnormal result.
Using the clinical characteristics of our cohort, we calculated the positive predictive value for each parameter singly and in combination. We used this data to weight each parameter and calculate a cumulative risk score, dividing children into low and high risk of abnormal glucose tolerance when tested by OGTT (Fig. 1). We then applied this risk score to our cohort: 36 children scored as high risk, which included all 13 children with abnormal glucose tolerance. The sensitivity of this scoring system in our cohort is 100%, specificity 57%, and positive predictive value 36%, i.e., three children need to be tested to identify one abnormal result. Prospective validation of this assessment tool is now underway in our clinic. We suggest that other units develop risk assessment tools for their populations so that screening at-risk children for glucose intolerance becomes more feasible and is not “a lost battle.”