Association Between Inflammatory Biomarkers and Maximum Thyroid Nodule Diameter: An Exploratory Analysis Using Conventional Statistical and Machine-Learning Approaches

The relationship between systemic inflammation and thyroid nodule size has received less attention than its role in malignancy. This retrospective study evaluated the associations of the neutrophil-to-lymphocyte (NLR), platelet-to-lymphocyte (PLR), lymphocyte-to-monocyte (LMR), monocyte-to-high-density lipoprotein cholesterol (MHR), and neutrophil-to-high-density lipoprotein cholesterol (NHR) ratios with maximum ultrasonographic thyroid nodule diameter. Of 279 screened records, 161 patients aged 18–70 years were analyzed. Spearman correlations with Holm correction constituted the primary analysis; multivariable linear regression and five machine-learning regression algorithms evaluated with repeated five-fold cross-validation (n = 155) were exploratory. In unadjusted analysis, LMR was the only index significantly correlated with nodule diameter (ρ = 0.230, 95% bootstrap CI: 0.069–0.384; Holm-adjusted p = 0.020). After adjustment for age, sex, body mass index, thyroid functional status, and HDL cholesterol, LMR remained associated with nodule diameter (β = 1.824 mm per unit; 95% CI: 0.171–3.478; n = 144), although the model explained little variance (R2 = 0.106). Random Forest achieved the lowest cross-validated root mean squared error (10.94 ± 1.55 mm; R2 = 0.116 ± 0.305), and LMR showed the highest permutation importance. These indices alone were insufficient for clinically reliable prediction of nodule diameter, and the weak LMR association should be regarded as hypothesis-generating.