Objectives There is evidence to suggest that alterations in skin health often co-occur with type 2 diabetes (T2D). However, the underlying mechanisms remain unclear. Despite their clinical relevance, skin conditions are often underdiagnosed in diabetes care. This study applied variable selection methods to identify key dermatological factors linked to T2D and examined their direct and indirect associations. Design Prospective birth cohort study. Participants This study analysed data from the Northern Finland Birth Cohort 1966, including 1906 participants (1024 women, 882 men) at ages 31 and 46 years. At age 46 years, the cohort participants, living in the region of Northern Finland, were invited to a general health clinical examination including the assessment of their T2D status and a dermatological examination. Methods Prior to performing a path analysis, we developed a data-driven procedure to select 47 possible skin-related variables for the model. The variable selection combined test-based methods (ie, a univariate analysis and a backward, forward and stepwise logistic regression) and penalty-based/machine learning methods (ie, Least Absolute Shrinkage and Selection Operator, Ridge, minimax convex penalty and Sparse Stepwise Regression).
Objectives There is evidence to suggest that alterations in skin health often co-occur with type 2 diabetes (T2D). However, the underlying mechanisms remain unclear. Despite their clinical relevance, skin conditions are often underdiagnosed in diabetes care. This study applied variable selection methods to identify key dermatological factors linked to T2D and examined their direct and indirect associations. Design Prospective birth cohort study. Participants This study analysed data from the Northern Finland Birth Cohort 1966, including 1906 participants (1024 women, 882 men) at ages 31 and 46 years. At age 46 years, the cohort participants, living in the region of Northern Finland, were invited to a general health clinical examination including the assessment of their T2D status and a dermatological examination. Methods Prior to performing a path analysis, we developed a data-driven procedure to select 47 possible skin-related variables for the model. The variable selection combined test-based methods (ie, a univariate analysis and a backward, forward and stepwise logistic regression) and penalty-based/machine learning methods (ie, Least Absolute Shrinkage and Selection Operator, Ridge, minimax convex penalty and Sparse Stepwise Regression). The selected skin-related variables were then included in a path analysis accounting for sex, body mass index (BMI), diet, education, depression, anxiety and sleep quality. Results Eight skin-related variables were selected and included in the path analysis. Psoriasis (OR 1.105, 95% CI 1.058 to 1.156) and Tinea pedis (OR 1.067, 95% CI 1.04 to 1.096) showed direct associations with T2D. Hyperhidrosis (OR 1.017, 95% CI 1.005 to 1.037) exhibited an indirect effect. Conclusions These results underscored the interconnectedness of dermatological conditions such as psoriasis and Tinea pedis with T2D. These findings suggest that certain skin conditions may serve as potential indicators of T2D. Early recognition and appropriate management of these conditions may help improve patients' quality of life.