[Prognostic research in endocrinology: design and common errors]

This article provides a conceptual overview of the methodology and design of prognostic research. It presents cohort studies as the most suitable design for prognostic investigations and discusses in detail the distinctions between traditional etio logical cohort studies and those primarily aimed at predicting outcomes. The article reviews examples of prognostic studies that focus on predicting gestational diabetes mellitus, renal failure in type 2 diabetes mellitus, osteoporotic fractures in postmenopausal women, outcomes of infertility treatment in women with elevated follicle-stimulating hormone levels, and survival in patients with adrenocortical carcinoma. The importance of adhering to the methodological standards outlined in the STROBE and TRIPOD guidelines is emphasized. It also highlights the need to avoid common design and analytical pitfalls, such as measuring predictors after the outcome has occurred or using ROC analysis without developing a multivari able prediction model. Particular attention is given to employing appropriate statistical methods, as a lack of understanding of these methods may lead to biased inferences and diminish the scientific value of the published research. Incorporating well-designed prognostic cohort studies and rigorous analytical approaches will enhance clinical practice and strengthen the evidence base in modern endocrinology.