Development of a metabolic syndrome prediction model using smartphone-derived digital anthropometry

Integrating metabolic syndrome (MetS) screening procedures into routine care remains challenging. Traditional anthropometric and body composition assessments, while useful, have drawbacks that limit their application. However, automated anthropometrics produced from smartphone scanning applications may offer a solution. This study aimed to determine whether smartphone-derived anthropometrics could effectively predict both MetS and its severity. A total of 281 participants underwent a MetS screening assessment to determine fasting blood pressure, lipids, glucose and waist circumference and completed a smartphone scanning assessment (MeThreeSixty^®) to collect digital anthropometrics. Actual MetS classification and MetS severity (MetS_index), a continuous estimate of MetS progression, were determined using MetS screening data. Then, least absolute shrinkage and selection operator regression was used to develop a new MetS_index prediction equation in a subset of participants (<i>n</i> 226), which was subsequently tested in the remaining participants (<i>n</i> 55), and MetS classification was predicted from the retained variables using logistic regression. The following equation was produced: <i>Smartphone-predicted MetS</i> _<i>index</i> : -0·8880 + 0·1493(<i>medication use = 1; 0 = no medication use</i>) + 0·0089(<i>weight</i>) + 0·0079(<i>bust circumf.</i>) + 0·0140 (<i>thigh circumf.</i>) - 0·6247(<i>appendage-to-trunk circumf. index</i>), where <i>medication use</i> includes medications for hypertension, dyslipidaemia or hyperglycaemia. The newly developed MetS_index prediction model demonstrated equivalence with actual MetS_index and revealed acceptable agreement (R²:0·72; root mean squared error: 0·42; se of the estimate: 0·22) when evaluated in the testing sample (<i>n</i> 55), although proportional bias was observed (<i>P</i> < 0·001). Smartphone-predicted MetS classification demonstrated acceptable diagnostic performance with an accuracy of 92·7 % and an AUC of 0·89. Smartphone scanning applications can accurately assess MetS prevalence and severity, presenting new possibilities for health screening beyond clinical environments.