Main Article Content


Objective: This paper describes an unsupervised Machine Learning approach to estimate the HOMA-IR cut-off identifying subjects at risk of insulin resistance in a given ethnic group, based on the clinical data of a representative sample. Methods: We apply the approach to clinical data of individuals of Arab ancestors obtained from a family study conducted in the city of Nizwa between January 2000 and December 2004. First, we identify HOMA-IR-correlated variables to which we apply our own clustering algorithm. Two clusters having the smallest overlap in their HOMA-IR values are returned. These clusters represent samples of two populations: insulin sensitive subjects and individuals at risk of insulin resistance. The cut-off value is estimated from intersections of the Gaussian functions modelling the HOMA-IR distributions of these populations. Results: We identified a HOMA-IR cut-off value of 1.62+/-0.06. We demonstrated the validity of this cut-off by 1) Showing that clinical characteristics of the identified groups match well published research findings about insulin resistance. 2) Showing a strong relationship between the segmentations resulting from the proposed cut-off and that resulting from the 2-hours glucose cut-off recommended by WHO for detecting prediabetes. Finally, we showed that the method is also able to identify cut-off values for similar problems (e.g. fasting sugar cut-off for prediabetes). Conclusion: The proposed method defines a HOMA-IR cut-off value for detecting individuals at risk of insulin resistance. Such method can identify high risk individuals at early stage which may prevent or at least delay the onset of chronic diseases like type 2 diabetes.

Keywords: Machine Learning; Feature Selection; K-mean++ Clustering; Insulin Resistance; HOMA-IR; T2DM.

Article Details

How to Cite
Abdessalem, A., Zidoum, H., Zadjali, F. ., Hedjam, R., Al-Ansari, A., Bayoumi, R., Al-Yahyaee, S., Hassan, M., & Albarwani, S. (2021). Estimate of the HOMA-IR Cut-off Value Identifying Subjects at Risk of Insulin Resistance using a Machine Learning Approach. Sultan Qaboos University Medical Journal [SQUMJ], 1(1).