Cytokine Signature and Nomogram Flag Interstitial Lung Disease in Treatment-Naïve Rheumatoid Arthritis

Plasma levels of the immune messenger interleukin-22 were significantly lower in rheumatoid arthritis patients whose lungs showed interstitial disease than in comparable patients without lung involvement, while tumor necrosis factor-alpha ran in the opposite direction, according to a retrospective case-cohort study of 66 treatment-naïve patients conducted at a single Chinese hospital. Building on those two cytokine signals, the researchers constructed a classification nomogram that combined age, mean corpuscular hemoglobin concentration, interleukin-22, log-transformed tumor necrosis factor-alpha, and smoking history. Under fivefold cross-validation, the model achieved an area under the receiver-operating characteristic curve of 0.851, with a 95% confidence interval of 0.755 to 0.947, indicating discriminative performance well above chance for distinguishing patients with rheumatoid arthritis-associated interstitial lung disease from those without it.

Rheumatoid arthritis is a chronic, progressive systemic autoimmune disorder characterized by synovial inflammation, joint destruction, and a range of extra-articular manifestations. Among these, interstitial lung disease stands out as a serious complication involving progressive pulmonary fibrosis that contributes substantially to mortality. The study’s introduction notes that rheumatoid arthritis-associated interstitial lung disease accounts for roughly 10 to 20% of mortality in these patients, yet the precise immunological mechanisms that predispose certain individuals to pulmonary fibrosis remain poorly understood, and reliable serum biomarkers for early detection are urgently needed. Prior work had hinted that interleukin-22 expression may be reduced in patients with the lung complication, potentially inversely correlating with the severity of pulmonary involvement, and murine models of pulmonary fibrosis had suggested a protective role for this cytokine against fibrotic change.

The cytokine results were clear for two of the nine molecules tested. Plasma interleukin-22 averaged 205.36 plus or minus 44.81 picograms per milliliter in the interstitial lung disease group versus 246.92 plus or minus 41.53 picograms per milliliter in the group without lung involvement, a difference reaching p less than 0.001. Tumor necrosis factor-alpha, reported as median with interquartile range because of its skewed distribution, was 2.43 with an interquartile range of 1.99 to 5.56 picograms per milliliter in the lung disease group against 1.65 with a range of 1.15 to 2.91 in the comparison group, at p equal to 0.012. No significant between-group differences emerged for interleukin-17, interleukin-2, interleukin-4, interleukin-6, interleukin-10, interleukin-12p70, or interferon-gamma. Because the cohort was older and contained more smokers in the lung disease arm, the analysts refit multivariable linear regression models with each cytokine as the dependent variable, adjusting for age and smoking history; interleukin-22 remained significantly lower and tumor necrosis factor-alpha significantly higher after adjustment.

Correlation analyses added texture to the picture. Moderate to strong positive correlations appeared among interleukin-6, interleukin-10, interleukin-12p70, interferon-gamma, and tumor necrosis factor-alpha. Interleukin-6, interleukin-10, and tumor necrosis factor-alpha each correlated positively with clinical disease activity measures including the Disease Activity Score 28, swollen joint count, and tender joint count. Tumor necrosis factor-alpha was positively correlated with the computed tomography fibrosis score, whereas interleukin-22 was negatively correlated with it. Partial Spearman correlation analyses controlling for age and smoking history confirmed these associations. The authors caution, however, that no formal adjustment for multiple comparisons was applied across the nine cytokines and multiple clinical variables tested, so the findings should be interpreted conservatively and confirmed in independent cohorts.

Demographically, the two arms differed in ways that mattered for modeling. Patients with interstitial lung disease were significantly older, averaging 66.15 plus or minus 10.40 years against 54.39 plus or minus 11.06 years, and 27.3% of them smoked compared with 6.1% of the others, both p less than 0.05. Sex distribution, hypertension, and coronary heart disease did not differ significantly. The lung disease group also showed higher Disease Activity Score 28 values, swollen joint counts, and tender joint counts, along with lower mean corpuscular hemoglobin concentration, all at p less than 0.05. Univariable logistic regression flagged age, smoking status, interleukin-22, mean corpuscular hemoglobin concentration, Disease Activity Score 28, swollen and tender joint counts, and tumor necrosis factor-alpha as significantly associated with the lung complication. Least absolute shrinkage and selection operator regression and Boruta feature selection then converged on five variables—age, mean corpuscular hemoglobin concentration, interleukin-22, natural log-transformed tumor necrosis factor-alpha, and smoking—for the final model, with smoking retained on the basis of its established clinical relevance even though Boruta did not select it.

The cross-validated area under the receiver-operating characteristic curve was 0.851 with a 95% confidence interval of 0.755 to 0.947, and a sensitivity analysis that dropped smoking still yielded a cross-validated area of 0.841, which the authors cite as evidence of robustness. The calibration plot showed acceptable agreement between predicted and observed probabilities, and decision curve analysis indicated that the nomogram provided greater net benefit than either treat-all or treat-none strategies across threshold probabilities from roughly 0.05 to 0.70, with the advantage fading at very high thresholds above 0.7. Higher age, lower mean corpuscular hemoglobin concentration, lower interleukin-22, elevated tumor necrosis factor-alpha, and smoking each push the total nomogram score upward, signaling a higher probability of interstitial lung disease.

The interpretation carries important qualifications. The authors emphasize that the nomogram is designed for classifying existing rheumatoid arthritis-associated interstitial lung disease rather than predicting future onset; as a cross-sectional analysis it identifies concurrent disease, and prospective validation would be needed to assess any predictive utility for incident cases. Because the 1:1 case-control design artificially fixes apparent prevalence at 50%, the predicted probabilities systematically overestimate absolute risk when applied to unselected clinical cohorts, where clinically significant interstitial lung disease affects roughly 10 to 18% of rheumatoid arthritis patients and an additional 19 to 67% show subclinical disease. The model is therefore intended for risk stratification and identification of high-risk individuals rather than absolute risk estimation, and recalibration of the model intercept with real-world prevalence data would be required before clinical use for that purpose. Its discriminative ability, the authors argue, remains valid for ranking individuals by relative risk.

Limitations frame the contribution as exploratory. The sample of 66 patients limited statistical power—possibly explaining why circulating interleukin-17A did not differ between groups despite mechanistic expectations from animal work—and raised overfitting risk, with an events-per-variable ratio of approximately 6.6 falling below the commonly recommended threshold of 10 and an initial candidate pool of roughly twenty variables adding further selection pressure. The retrospective, single-center design introduces potential selection bias and limits generalizability, cytokine measurements from peripheral blood may not reflect the local lung microenvironment, inter-observer reliability of the fibrosis score was not formally assessed, and established serological markers such as rheumatoid factor and anti-CCP antibodies were absent from the model despite being well-recognized correlates of the lung complication. The restriction to newly diagnosed, treatment-naïve patients minimizes drug-related confounding but may limit applicability to patients on ongoing treatment. The authors call for external validation in larger, prospective, multi-center cohorts, longitudinal studies of whether baseline interleukin-22 and tumor necrosis factor-alpha predict future development of interstitial lung disease, and exploration of cytokine profiles in bronchoalveolar lavage fluid. Within those bounds, the study offers a hypothesis-generating tool built from routinely available parameters—age, a red cell index, two plasma cytokines, and smoking history—that together outperform single biomarkers for flagging which rheumatoid arthritis patients may warrant closer lung imaging and monitoring.

Subject of Research: Cytokine profiles and a classification nomogram for interstitial lung disease in treatment-naïve rheumatoid arthritis patients.

Article Title: Analysis of Factors Associated With Interstitial Lung Disease in Treatment‐Naïve Rheumatoid Arthritis Patients and Development of a Classification Nomogram

Article References: Lu, L., Chen, Y., Zhang, T., Cao, L., Ma, H., Song, Y., & Zhou, D. (2026). Analysis of Factors Associated With Interstitial Lung Disease in Treatment‐Naïve Rheumatoid Arthritis Patients and Development of a Classification Nomogram. Immunity, Inflammation and Disease, 14(10), Article e70571. https://doi.org/10.1002/iid3.70571

Image Credits: AI Generated

DOI: 10.1002/iid3.70571

Keywords: rheumatoid arthritis, interstitial lung disease, interleukin-22, tumor necrosis factor-alpha, nomogram, cross-validation, high-resolution computed tomography, pulmonary fibrosis, cytokines, biomarkers, treatment-naïve patients, risk stratification