Machine Learning Predicts Hyperglycemia Risk in Psoriasis

Researchers have developed and validated a machine learning model using XGBoost to predict hyperglycemia risk in patients with psoriasis.¹ Tested on both clinical and National Health and Nutrition Examination Survey (NHANES) data sets, the model demonstrated strong accuracy and clinical utility, offering a web-based calculator to guide personalized treatment strategies for high-risk individuals.

This retrospective study is published in Clinical, Cosmetic, and Investigational Dermatology.

“As of right now, there is no prediction model for psoriasis patients’ risk of hyperglycemia,” wrote the researchers of the study. “Therefore, in order to screen out the high-risk group and implement early intervention, as well as to provide guidance for the individualized treatment of psoriasis patients, we propose to build a prediction model for the risk of hyperglycemia in psoriasis patients using clinical data of psoriasis patients and machine learning algorithms.”

Psoriasis is a condition that is linked to systemic inflammation and can impair glucose metabolism, leading to insulin resistance and hyperglycemia.² Elevated stress hyperglycemia in patients with psoriasis has also been associated with worse outcomes, including higher all-cause mortality.

This study collected clinical data from 575 patients with psoriasis treated in the dermatology department at the Affiliated Hospital of Guilin Medical University in China, who were randomly divided into a training set (70%) and an internal test set (30%).¹ An external validation set included 135 patients from the NHANES cohorts of 2003-2004 and 2011-2012. Eleven machine learning algorithms were systematically trained and compared, and model performance was assessed to evaluate both predictive accuracy and clinical utility.

The XGBoost model was selected as the final predictive algorithm due to its superior performance across multiple metrics. The model achieved an area under the curve (AUC) of 0.821 (95% CI, 0.775-0.866) in the training set, 0.820 (95% CI, 0.751-0.888) in the internal test set, and 0.788 (95% CI, 0.695-0.881) in the external NHANES test set, demonstrating consistent accuracy across datasets.

Additionally, calibration curves indicated strong agreement between predicted and observed hyperglycemia risk, while clinical decision curve analysis confirmed its potential utility for guiding treatment decisions. Furthermore, a web-based calculator was developed to make the model easily accessible for clinicians managing patients with psoriasis at risk for hyperglycemia.

However, the researchers noted some limitations. First, the data came from a single hospital in China, and population differences may have introduced bias. Additionally, the retrospective design lacked key clinical indicators, which limited direct therapeutic use. Multicenter prospective studies are needed to validate the model.

Despite these limitations, the researchers believe the XGBoost model effectively predicts hyperglycemia risk in patients with psoriasis, supporting personalized management strategies. With further validation in diverse populations, this tool could help clinicians identify high-risk individuals and guide targeted interventions.

“Since hyperglycemia and psoriasis are linked, we developed a prediction model for the incidence of hyperglycemia in psoriasis patients using their clinical examination indexes and general data,” wrote the researchers. “We also investigated the relationship between the 2 in a preliminary manner, which gave guidance for the next study. Five of these indicators—age, BUN [blood urea nitrogen], ALT [alanine aminotransferase], HDL-C [high-density lipoprotein cholesterol], and TG [triglycerides]—may have a rather strong correlation with blood glucose. A customized treatment plan is required for psoriasis patients who are also hyperglycemic or at risk for hyperglycemia in order to coordinate the management of psoriasis inflammation and glycemic development.”

References

1. Hu M, Chen D, Yu J. Predictive modeling of the risk of hyperglycemia in psoriasis patients using machine learning: a multicenter retrospective study. Clin Cosmet Investig Dermatol. 2025;18:3667-3680. doi:10.2147/CCID.S552796

2. Steinzor P. Stress hyperglycemia patio linked with all-cause mortality in patients with psoriasis. AJMC^®. November 7, 2024. Accessed January 5, 2026. https://www.ajmc.com/view/stress-hyperglycemia-ratio-linked-with-all-cause-mortality-in-patients-with-psoriasis