Spotlight on racial differences in risk assessment

Two recent publications looked at the issue of race in type 2 diabetes prediction.

A new scientific statement from the American Heart Association offers clinicians data and guidance on the epidemiology of diabetes and atherosclerotic cardiovascular disease (ASCVD) among Asian American people. It notes that prevalence of type 2 diabetes is higher in all Asian American subgroups compared with non-Hispanic White people and that risk for ASCVD is highest among South Asian and Filipino adults but lowest among Chinese, Japanese, and Korean adults.

The statement advises clinicians to have a lower threshold for diagnosing overweight or obesity in Asian American patients: body mass index cutoffs of 23 kg/m² and 27.5 kg/m², respectively. Regarding treatment, lifestyle interventions should be tailored to patients' cultures, and clinicians should be cautious about the potential for drug interactions given common use of complementary and alternative medicines and self-prescription in some populations. The statement notes that it is challenging to provide more evidence-based recommendations due to limited data on Asian American patients in risk prediction models, national surveillance surveys, and clinical trials. Finally, the statement, which was published by Circulation on May 8, urges those who care for Asian American patients to “become more involved in promoting community-based health promotion efforts.”

A study published by PLOS Global Public Health on May 17 looked at the algorithms for predicting type 2 diabetes and found that they may be racially biased. The researchers used National Health and Nutrition Examination Survey data (including fasting blood tests) on 9,987 adults without a prior diagnosis of diabetes, gathered between 1999 and 2010, to assess the Prediabetes Risk Test (PRT) issued by the National Diabetes Prevention Program, the Framingham Offspring Risk Score, and the Atherosclerosis Risk in Communities (ARIC) Study Model. They found all the investigated models to be miscalibrated with regard to race consistently across the survey years. The Framingham Offspring Risk Score overestimated type 2 diabetes risk for non-Hispanic White patients and underestimated risk for non-Hispanic Black patients. The PRT and the ARIC models overestimated risk for both White and Black patients, but more so for White patients. The authors note that while under- and overestimates both carry risks, the interventions for prediabetes are low risk enough that underestimates are likely more problematic. They suggested that the inaccuracies may be due to the “absence of socio-economic determinants in the models such as health literacy status and access to healthcare,” and they called for prediction models to “demonstrate algorithmic fairness before adoption.”