Laboratory parameters-based logistic regression models for rapid screening of thyroid nodules.

Abstract

The increasing incidence of thyroid nodules (TNs) are placing mounting pressure on radiologists. Our study aimed to evaluate the effectiveness of laboratory parameters in the detection of benign and malignant TNs and develop early diagnosis logistic regression models by using the laboratory parameters. This study was conducted from December 2016 to July 2022 at Beijing Chaoyang Hospital. Totals of 251 healthy individuals, 176 patients with benign TNs (BTNs), and 302 patients with malignant TNs (MTNs) were enrolled. Univariate and multivariate logistic regression analyses were performed to find the meaningful laboratory factors of TNs, and subsequently, prediction models were established. Sensitivity, specificity, and receiver operating characteristic (ROC) curve analysis were applied to evaluate the predictive value of the regression equations. We also compared the expression levels of meaningful indexes in different types of individuals. The models were verified by the validation cohort. Based on the meaningful laboratory factors selected by regression analysis, for predicting patients with BTNs and MTNs in healthy individuals, the diagnostic models were Logit(P) = -2.525 × high density lipoprotein cholesterol (HDL-C) + 1.515 × glucose (Glu) + 0.003 × total triiodothyronine (TT3) - 4.607 × free triiodothyronine (FT3) - 0.81 × serum thyroid stimulating hormone (sTSH) + 8.585 and Logit(P) = -2.789 × HDL-C + 0.035 × lipoprotein [Lp(a)] + 1.141 × Glu + 0.054 × antithyroglobulin antibody (Anti-Tg) - 1.931 × FT3 - 0.341 × sTSH + 3.757. Ideally, the two models showed high area under the curve (AUC) values. For distinguishing patients with BTNs and MTNs, the diagnostic model was Logit(P) = -0.303 × Glu + 0.335 × sTSH + 1.535. However, this model had a relatively low AUC. Our research shows that TNs are associated with laboratory indexes about metabolism of Glu and lipid, thyroid function, albumin (ALB), mean corpuscular hemoglobin (MCH), and platelet (PLT). In routine physical examination and early screening of TNs, laboratory parameters-based logistic regression models are recommended.