Prognostic Value of Clinical and FDG-PET/CT-Derived Parameters for Overall Survival In Oligometastatic Non–Small Cell Lung Cancer Patients
Abstract
Purpose
Prognostic stratification of patients with oligometastatic (1–5 metastases) non–small cell lung cancer (omNSCLC) remains a challenging yet important task. This study evaluates the prognostic value of clinical and FDG-PET-derived parameters for predicting overall survival (OS), measured from the diagnosis of oligometastatic disease.
Methods
A retrospective cohort of 160 omNSCLC patients was analyzed. Clinical and imaging-derived parameters included patient age, sex, treatment modality, metastatic presentation pattern, total tumor volume, standardised uptake value (SUV), total lesion glycolysis, organ involvement pattern, and radiomics. OS was analyzed using Kaplan–Meier curves with log-rank testing (α=0.05). Hazard ratios (HRs) were calculated using univariate Cox proportional hazards regression (α=0.05). The multivariable Cox model was tested with repeated 5-fold cross-validation and evaluated with concordance index (C-index), Brier score, and time-dependent AUC over 1–5 years of follow-up.
Results
Kaplan–Meier analysis identified several variables that stratified patients into distinct OS risk groups. A total tumor volume >17.6 mL, SUVmax >13.1 g/mL, presence of more than one distant metastasis, and abdominal involvement were each associated with significantly poorer OS (p≤0.036). Univariate Cox regression confirmed their relevance, yielding HRs of 1.7–1.9 (p≤0.006, C-index 0.55–0.60). The multivariable Cox model demonstrated only moderate predictive performance, with a C-index of 0.64 (95% CI 0.63–0.65), an AUC of 0.69 (95% CI 0.67–0.70), and a Brier score of 0.21 (95% CI 0.21–0.22). The final model included age (β=0.02), time from primary to OMD diagnosis (years; β=-0.05), log-transformed lesion volume (β=0.14), number of metastases (β=0.31), presence of abdominal lesions (β=0.33), and exclusive thoracic involvement (β=-0.06). Inclusion of radiomics features did not improve the model’s performance.
Conclusion
While conventional clinical and imaging-derived parameters enable meaningful OS stratification in omNSCLC patients, accurate multivariable survival prediction remains challenging. Radiomics features did not add prognostic value, highlighting the need for cautious integration of high-dimensional imaging biomarkers and further external validation.