Objective To investigate the associations of non-invasive peripheral blood inflammatory indices, cytokines, and lymphocyte subsets with the risk of rheumatoid arthritis (RA)-associated interstitial lung disease (RA-ILD), and to evaluate the auxiliary diagnostic discrimination of key biomarker panels. Methods This retrospective study enrolled 216 eligible patients (148 with RA alone, 68 with RA-ILD). After 1:1 propensity score matching (PSM), 114 patients were included. Variables were screened using the least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression. A nomogram was developed and internally validated, with discriminative performance and calibration assessed by the area under the receiver operating characteristic curve (AUC), calibration curves, and bootstrap resampling. An exploratory ROC-derived cutoff of the model-predicted probability was further determined using the Youden index. An (extreme gradient boosting) XGBoost model was further constructed and combined with Shapley Additive exPlanations (SHAP) analysis for performance comparison and interpretability evaluation. Results In multivariable analysis, CD4⁺ T cells, the systemic inflammation response index (SIRI), and the interleukin (IL)-2/IL-6 ratio were independently associated with RA-ILD. Restricted cubic spline (RCS) analyses confirmed that CD4⁺ T cells and the IL-2/IL-6 ratio exhibited non-linear correlations with RA-ILD risk, whereas SIRI showed an approximately linear positive trend. The nomogram achieved an AUC of 0.820, compared with 0.947 for the XGBoost model. A model-predicted probability cutoff of 0.494, corresponding to a nomogram score of 111.25 points, was associated with higher odds of RA-ILD (OR = 10.40, 95% CI: 4.51–25.59). SHAP analysis identified the IL-2/IL-6 ratio, SIRI, IFN-γ/IL-4 ratio, IL-17A, and CD4⁺ T cells as the leading contributors to model output. Conclusion Peripheral blood immune and inflammatory biomarkers are closely related to RA-ILD risk. The combination of peripheral blood immune and inflammatory biomarkers shows favorable discriminatory value and may provide a laboratory-based reference for auxiliary identification of RA-ILD. <Table Float="No" ID="Taba"> <tgroup cols="2"> <colspec align="left" colname="c1" colnum="1" /> <colspec align="left" colname="c2" colnum="2" /> <tbody> <row> <entry align="left" nameend="c2" namest="c1"> Key Points • This study explored RA-ILD from a multidimensional peripheral blood immune-inflammatory perspective. • CD4⁺ T cells, SIRI, and the IL-2/IL-6 ratio emerged as key variables in the multivariable analysis. • CD4⁺ T cells and the IL-2/IL-6 ratio showed non-linear associations with RA-ILD risk, whereas SIRI showed an approximately linear positive association. • The proposed models showed potential for RA-ILD risk assessment and warrant further validation. </entry> </row> </tbody> </tgroup> </Table>

Multidimensional immunoinflammatory profiles associated with rheumatoid arthritis–associated interstitial lung disease: a cross-sectional exploratory study

期刊论文
发表日期：

2026年6月20日

查看全文

Yanchun Zhao,
Ziyun Guan,
Limei Peng,
Ruyi Zhong,
Minying Liu,
Xueqing Chen

摘要

Objective

To investigate the associations of non-invasive peripheral blood inflammatory indices, cytokines, and lymphocyte subsets with the risk of rheumatoid arthritis (RA)-associated interstitial lung disease (RA-ILD), and to evaluate the auxiliary diagnostic discrimination of key biomarker panels.

Methods

This retrospective study enrolled 216 eligible patients (148 with RA alone, 68 with RA-ILD). After 1:1 propensity score matching (PSM), 114 patients were included. Variables were screened using the least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression. A nomogram was developed and internally validated, with discriminative performance and calibration assessed by the area under the receiver operating characteristic curve (AUC), calibration curves, and bootstrap resampling. An exploratory ROC-derived cutoff of the model-predicted probability was further determined using the Youden index. An (extreme gradient boosting) XGBoost model was further constructed and combined with Shapley Additive exPlanations (SHAP) analysis for performance comparison and interpretability evaluation.

Results

In multivariable analysis, CD4⁺ T cells, the systemic inflammation response index (SIRI), and the interleukin (IL)-2/IL-6 ratio were independently associated with RA-ILD. Restricted cubic spline (RCS) analyses confirmed that CD4⁺ T cells and the IL-2/IL-6 ratio exhibited non-linear correlations with RA-ILD risk, whereas SIRI showed an approximately linear positive trend. The nomogram achieved an AUC of 0.820, compared with 0.947 for the XGBoost model. A model-predicted probability cutoff of 0.494, corresponding to a nomogram score of 111.25 points, was associated with higher odds of RA-ILD (OR = 10.40, 95% CI: 4.51–25.59). SHAP analysis identified the IL-2/IL-6 ratio, SIRI, IFN-γ/IL-4 ratio, IL-17A, and CD4⁺ T cells as the leading contributors to model output.

Conclusion

Peripheral blood immune and inflammatory biomarkers are closely related to RA-ILD risk. The combination of peripheral blood immune and inflammatory biomarkers shows favorable discriminatory value and may provide a laboratory-based reference for auxiliary identification of RA-ILD.

Key Points

• This study explored RA-ILD from a multidimensional peripheral blood immune-inflammatory perspective.

• CD4⁺ T cells, SIRI, and the IL-2/IL-6 ratio emerged as key variables in the multivariable analysis.

• CD4⁺ T cells and the IL-2/IL-6 ratio showed non-linear associations with RA-ILD risk, whereas SIRI showed an approximately linear positive association.

• The proposed models showed potential for RA-ILD risk assessment and warrant further validation.