Diagnostic and prognostic value of the injury–inflammation–fibrosis serum biomarker panel (KL-6, SAA, YKL-40) in lung cancer treatment-associated ILD

OBJECTIVE: Cancer therapy-related interstitial lung disease (CT‑ILD) is a serious and potentially fatal complication in lung cancer patients treated with radiotherapy or antineoplastic agents, characterized by high mortality and limited treatment options. Early recognition remains challenging due to the absence of sensitive and specific biomarkers. This study aimed to establish a tripartite serum biomarker panel-KL‑6 (lung injury), SAA (inflammation), and YKL‑40 (fibrosis)-representing the "injury-inflammation-fibroproliferation" axis, and to evaluate its diagnostic and prognostic value for CT‑ILD. METHODS: In this single-center observational case-control study, we consecutively enrolled hospitalized patients with histologically confirmed lung cancer who underwent thoracic radiotherapy and/or systemic anti-neoplastic therapy between January 2024 and June 2025. Based on imaging findings, patients were classified into the ILD group (n = 73) or N‑ILD group (n = 67). Serum KL‑6, SAA, and YKL‑40 were quantified alongside conventional inflammatory markers (CRP, IL‑6). ROC analysis evaluated diagnostic performance. Logistic regression identified independent ILD correlates. Subgroup analyses delineated etiological differences, comparing radiotherapy-induced ILD (RT‑ILD) with drug-induced ILD (DI‑ILD), and further explored their associations with clinical benefit as an exploratory endpoint. RESULTS: KL‑6, SAA, YKL‑40, and IL‑6 were significantly elevated in ILD compared with N‑ILD (all P < 0.05). ROC analysis yielded AUCs of 0.764 for KL‑6, 0.880 for SAA, and 0.718 for YKL‑40, with optimal cut‑offs of 349.5 U/L, 18.7 mg/L, and 71.65 ng/mL, respectively (all P < 0.001). Independent correlates of ILD included SAA > 18.7 mg/L (OR = 32.321), KL‑6 > 349.5 U/L (OR = 5.842), and YKL‑40 > 71.65 ng/mL (OR = 5.250). In the overall cohort, the combined model (KL-6 + SAA + YKL-40) showed good calibration (Hosmer-Lemeshow P > 0.05) but did not significantly outperform SAA alone in AUC (DeLong test, P > 0.05). SAA maintained high diagnostic value in both RT-ILD (AUC = 0.834) and DI-ILD (AUC = 0.891) subgroups. In exploratory analyses, higher KL-6 levels and the presence of multiple metastases were associated with a lack of durable clinical benefit (both P < 0.05). CONCLUSION: The KL‑6/SAA/YKL‑40 biomarker panel demonstrates exploratory utility for early CT‑ILD detection. SAA alone shows high diagnostic performance, while the combined model offers acceptable calibration. The panel may support risk stratification and timely individualized management during chemo‑/radiotherapy.