Current Web-Based Risk Tool May Overestimate Risk of Pre-Diabetes Interview with:

Dr. Saeid Shahraz Assistant Professor of Medicine Tufts Medical Center

Dr. Saeid Shahraz

Dr. Saeid Shahraz
Assistant Professor of Medicine
Tufts Medical Center What is the background for this study?
Response: American Diabetes Association (ADA) has set up a lower cut point for diagnosing prediabetes ( those with Impaired Fasting  Glucose   100 mg/dL) compared to the World Health Organization’s cut point, which is 110 mg/dL. This arbitrariness in cut point definition triples the number of cases labeled as prediabetes.

Along with lowering the diagnostic threshold by the ADA, the Centers for Disease Control and Prevention (CDC), the American Medical Association (AMA), and the ADA endorsed and advertised a web-based risk model to define high-risk population for prediabetes. The risk engine asks a few questions ( age, sex, family history of diabetes, history of gestational diabetes and high blood pressure, physical activity and weight) and outputs a score that defines if the person is at risk for prediabetes. We suspected that the risk engine might overestimate the risk. What are the main findings?

Response:  3 out of 5 people 40 years or older and 8 out of 10 people 60 or older are at high risk for prediabetes. What should readers take away from your report?

Response: For people: If the risk engine, classifies them as being high-risk for prediabetes, they should not worry that much about the risk, as it might be a false result. In other words, there is a good chance that their blood sugar result shows normal.

For professional decision makers and researchers: it’s important to specify and calibrate a risk model that reasonably classifies populations for a specific medical condition. What recommendations do you have for future research as a result of this study?

Response: Testing the sensitivity and specificity of widely-advertised risk models that are shared with general public provides scientific evidence that helps understand the usefulness of the risk models and prevents from over-relying on their results. Is there anything else you would like to add?

Response: Our study may have limitations too. We used variables extracted from a population survey that represents the US population. We selected the variables to be semantically as close as possible to those employed in the risk model. Our results, however, must be quite reliable even with a high margin of error. Thank you for your contribution to the community.


Note: Content is Not intended as medical advice. Please consult your health care provider regarding your specific medical condition and questions.

More Medical Research Interviews on

[wysija_form id=”5″]