Abstract
Several inflammation scores have shown association with survival outcomes for patients with neuroendocrine tumours (NET) treated with peptide receptor radionuclide therapy (PRRT). However, whether these scores add value to established prognostic factors remains unknown. In this retrospective, cohort study of 557 NET patients undergoing PRRT in a tertiary referral centre from 2005 to 2015, we examined inflammatory markers and scores previously associated with cancer outcomes, using Cox proportional hazard models and Akaike's information criterion. Lower albumin (hazard ratio [95% confidence interval], .91 [.87-.95] per unit), as well as higher C-reactive protein (CRP; 1.02 [1.01-1.02]), Glasgow Prognostic Score (GPS; 1 vs. 0: 1.67 [1.14-2.44], 2 vs. 0 3.60 [2.24-5.79]), CRP/albumin ratio (1.84 [1.43-2.37]) and platelet count (Plt) × CRP, but not white blood cell, neutrophil and thrombocyte counts or derived neutrophil to lymphocyte ratio (dNLR), were associated with shorter median overall survival (OS) in an adjusted analysis. The addition of parameters based on albumin and CRP, but not dNLR, to a base model including age, chromogranin A, the cell proliferation marker Ki-67, performance status, tumour site and previous treatments improved the predictive accuracy of the base model. In an exploratory analysis of patients with available erythrocyte sedimentation rate (ESR) and CRP, ESR emerged as the most powerful predictor. When added to a prognostic model for OS in NET patients treated with PRRT, most inflammation scores further improved the model. Albumin was the single marker adding most value to the set of established prognostic markers, whereas dNLR did not seem to improve the model's prognostic ability.