Comparison of prognostic models for predicting severe noninfectious complications in common variable immunodeficiency

A significant clinical heterogeneity of common variable immunodeficiency (CVID) complicates the prediction of severe noninfectious complications (NICs), thereby hindering personalized treatment strategies. This study aimed to compare retrospectively the performance of three distinct prognostic models-low switched memory B-cell (SMB) count, the composite Variable Immunodeficiency Score by Upfront Analytic Link (VISUAL), and the monogenic common variable immunodeficiency (Mo-CVID) score-for predicting NIC development in a single-center Turkish patient cohort. Clinical and immunologic data from 71 CVID patients were analyzed. Univariate analyses were used to compare patient groups, while multivariate logistic regression was employed to identify independent risk factors. During follow-up, 62.0% of patients developed at least one NIC. At diagnosis, patients with NICs exhibited significant immunologic differences, including lower platelet counts, a reduced CD4-CD8 ratio, and significantly lower SMB proportions. The multivariate analysis revealed that only the VISUAL score was a significant and independent predictor of NICs (odds ratio: 1.481; P = 0.005). Furthermore, receiver operating characteristic analysis confirmed its good discriminatory ability (area under the curve: 0.764), with an optimal cut-off value of >8.5. In contrast, neither a low SMB count as a categoric variable nor the Mo-CVID score remained as independent predictors in the final model. The VISUAL score, which integrates multiple B- and T-cell abnormalities, is a superior and more reliable tool for risk stratification in CVID. Its successful validation supports its use as a universal clinical instrument to identify high-risk individuals who may benefit from closer surveillance and earlier therapeutic interventions.