C‐reactive protein flare predicts response to checkpoint inhibitor treatment in melanoma

Background The treatment of melanoma has been revolutionized by the use of immune checkpoint inhibition (ICI), but many patients do not benefit. Furthermore, immune‐related adverse events may occur during therapy. A predictive biomarker is needed to reliably identify patients benefitting. In lung, renal cell and bladder cancer early C‐reactive protein (CRP) kinetics were shown to be a predictive biomarker for ICI. Objective Here, we investigate early CRP kinetics as predictive biomarker for ICI in melanoma patients. Methods Two independent prospectively collected cohorts were analysed: Cohort 1 (n = 87) with advanced and Cohort 2 (n = 99) with completely resected melanoma. Patients were stratified by in the dynamics of CRP after ICI initiation: A doubling of baseline CRP within 30 days followed by at least a 30% drop within 3 months was classified as a CRP flare. If no doubling of CRP was reported, but a 30% drop within 3 months, patients were classified as CRP responders and all others as CRP non‐responders. Analysed factors included clinical characteristics like S100B and LDH. Median follow‐up was 1.5 and 1.7 years for Cohorts 1 and 2. Results In Cohort 1 CRP flare (n = 12), CRP responders (n = 43) and CRP non‐responders (n = 32) with a progression‐free survival (PFS) of 0.7, 0.6 and 0.2 years (p = 0.017) and an overall survival (OS) of 2.2, 1.5 and 1.0 years (p = 0.014), respectively. Multivariable Cox analysis showed an independent risk reduction of progression for CRP responders by 62% compared to CRP non‐responders (p = 0.001). In Cohort 2 CRP flare (n = 13), CRP responders (n = 70) and CRP non‐responders (n = 16) the log‐rank analysis showed a significant difference between OS and recurrence‐free survival (RFS) curves (p = 0.046 and p = 0.049). Conclusion Early CRP kinetics could indicate a response to ICI with improved OS and RFS/PFS. CRP flare and CRP response indicating significantly improved outcomes compared to CRP non‐responders.