A risk score for identifying overweight adolescents with dysglycemia in primary care settings1)

Abstract

Objective To develop a clinical risk scoring system for identifying adolescents with dysglycemia (prediabetes or diabetes) who need further confirmatory testing and to determine whether the addition of non-fasting tests would improve the prediction of dysglycemia. Study Design A sample of 176 overweight and obese adolescents (10 – 17 years) had a history/physical exam, a 2-h oral glucose tolerance test, and non-fasting tests [hemoglobin A1c, 1-h glucose challenge test (GCT), and random glucose test] performed. Given the low number of children with diabetes, we created several risk scoring systems combining the clinical characteristics with non-fasting tests for identifying adolescents with dysglycemia and compared the test performance. Results Sixty percent of participants were white and 32% were black; 39.2% had prediabetes and 1.1% had diabetes. A basic model including demographics, body mass index percentile, family history of diabetes, and acanthosis nigricans had reasonable test performance [area under the curve (AUC), 0.75; 95% confidence interval (95% CI), 0.68 – 0.82]. The addition of random glucose (AUC, 0.81; 95% CI, 0.75 – 0.87) or 1-h GCT (AUC, 0.82; 95% CI, 0.75 – 0.88) to the basic model significantly improved the predictive capacity, but the addition of hemoglobin A1c did not (AUC, 0.76; 95% CI, 0.68 – 0.83). The clinical score thresholds to consider for the basic plus random glucose model are total score cutoffs of 60 or 65 (sensitivity 86% and 65% and specificity 60% and 78%, respectively) and for the basic plus 1-h GCT model are total score cutoffs of 50 or 55 (sensitivity 87% and 73% and specificity 59% and 76%, respectively). Conclusions Pending a validation in additional populations, a risk score combining the clinical characteristics with non-fasting test results may be a useful tool for identifying children with dysglycemia in the primary care setting.

Addressing Prediabetes in Childhood Obesity Treatment Programs: Support from Research and Current Practice

“…38,50 An analysis of laboratory data from 1156 overweight and obese adolescents from National Health and Nutrition Examination Survey 1999-2006 found that an A1c cutoff of 6.5% was 75% sensitive for identifying diabetes, compared to a single FPG, though this study was limited by a very low number of undiagnosed diabetics (n = 4) and the possibility of erroneous reports of fasting. 51 The same study found that, for adolescents compared to adults, A1c was a significantly worse predictor of prediabetes defined by FPG (area under the receiver operating curve, 0.61 vs. 0.74; p < 0.01). A recent position statement from the Pediatric Endocrine Society notes that the diabetes diagnostic criteria are based on long-term health outcomes assessed in adults, but have not been validated in pediatric populations.…”

Predictors of Metabolic Risk in Childhood Obesity

“…Even if longitudinal studies on the exact predictive value of a family history of T2DM among obese children and adolescents are not available, cross-sectional data demonstrate that obese children and adolescents with a family history of T2DM are more likely to present with glucose dysregulation [38,39], suggesting an increased risk of adult T2DM in this sub-group of obese paediatric patients. Moreover, a very high percentage (74-100%) of children with T2DM has a family history of T2DM, proving that the absence of family history is a good negative predictor of paediatric-onset T2DM [40].…”

“…The American Diabetes Association's criteria for selecting children to test for T2DM and criteria specifically designed for screening obese children for IGT in European populations, such as a clinical score issued from a German cohort or the fasting glucose cut-off of 86 mg/dl issued from an Italian cohort [56,57], have recently shown sub-optimal accuracy in identifying IGT when tested in different populations [58]. Also, a clinical score based on data from a multi-ethnic American cohort showed only moderate accuracy [39]. Recently, fasting TG >103 mg/dl has been proposed as a criterion for identifying obese youth at risk of IGT in a Canadian population [58].…”

Estimating youth diabetes risk using NHANES data and machine learning

“…Studies focused on assessing youth preDM/DM risk to date have relied on relatively small sample sizes from localized clinical settings, and have sometimes included invasive blood tests that may not be the best initial strategy to assess risk. (45, 46) In contrast, NHANES includes a large sample of individuals from across the United States, including well-represented age, sex, and racial/ethnic subgroups, as well as detailed biomarker, clinical, and behavioral health data. While NHANES data have been used to develop diabetes risk screeners for adults,(25, 47, 48) and to examine prevalence of preDM/DM among youth,(6, 49) no studies before ours have used these data to develop and evaluate youth diabetes risk screeners.…”