Low albumin-to-globulin ratio is associated with interstitial lung disease in primary Sjögren's syndrome: a single-center retrospective cohort study with propensity score matching.

To investigate the association between albumin-to-globulin ratio (AGR) and interstitial lung disease (ILD) in patients with primary Sjögren's syndrome (pSS), to identify high-risk clinical phenotypes for SS-ILD, and to construct an AGR-based nomogram risk stratification model. This was a single-center retrospective cohort study enrolling pSS patients who met the 2016 ACR/EULAR classification criteria between January 2014 and December 2024. Patients were divided into ILD and non-ILD groups based on the presence or absence of ILD. Propensity score matching (PSM) was applied to balance baseline characteristics. Univariable and multivariable logistic regression analyses were performed to evaluate the independent association between AGR and ILD. A nomogram risk stratification model was constructed based on predictors identified in the multivariable analysis, and its predictive performance was assessed using receiver operating characteristic (ROC) curve analysis. A total of 1,087 pSS patients were enrolled (93.6% female; mean age 58.25 ± 12.16 years; ILD group n = 69; non-ILD group n = 1,018). In the total cohort, AGR was significantly lower in the ILD group than in the non-ILD group (1.01 ± 0.30 vs. 1.18 ± 0.31), and the prevalence of documented dry eye was markedly lower in the ILD group (2.9% vs. 19.4%). Multivariable regression analysis demonstrated that AGR was independently and inversely associated with ILD risk (OR = 0.138, 95% CI 0.038-0.502); CRP was also an independent risk factor (OR = 1.009, 95% CI 1.000-1.017); and absence of documented dry eye symptoms was independently associated with ILD (OR = 0.196, 95% CI 0.047-0.821). After PSM (1:2 nearest-neighbor matching on age, sex, and BMI), 67 of 69 ILD patients were successfully matched, yielding a balanced cohort of 201 patients (ILD n = 67; non-ILD n = 134), with all covariate standardized mean differences (SMDs) < 0.1. In the matched cohort, AGR remained significantly lower in the ILD group (1.01 ± 0.30 vs. 1.19 ± 0.30) and these associations remained robust in the multivariable model (AGR: OR = 0.127, 95% CI 0.024-0.687; absence of dry eye: OR = 0.145, 95% CI 0.029-0.720). Three layers of sensitivity analysis (basic IPTW, extended-covariate IPTW with 11 covariates, and E-value analysis) confirmed the robustness of the AGR-ILD association, with the E-value point estimate of 12.30 substantially exceeding the strength of known confounders. The exploratory nomogram constructed from AGR, dry eye, age, and LDH demonstrated an AUC of 0.710 (95% CI 0.647-0.774) at an optimal cutoff of 0.096, with a sensitivity of 54.5% (95% CI 42.4%-66.7%), specificity of 81.3% (95% CI 79.0%-83.6%), PPV of 16.6% (95% CI 12.9%-20.2%), NPV of 96.3% (95% CI 95.4%-97.2%), and accuracy of 79.6% (95% CI 77.2%-81.9%). In this single-center retrospective cohort, low AGR was independently associated with the occurrence of ILD in pSS patients. Low AGR combined with absence of documented dry eye symptoms may identify a clinical subset warranting further evaluation for pulmonary involvement. Given the retrospective design, limited ILD event count, and lack of external validation, the nomogram developed in this study should be considered an exploratory tool, and the findings are hypothesis-generating, requiring prospective multicenter validation before clinical application.