C‐reactive protein flare‐response predicts long‐term efficacy to first‐line anti‐PD‐1‐based combination therapy in metastatic renal cell carcinoma

Abstract Objectives Immune checkpoint blockade (IO) has revolutionised the treatment of metastatic renal cell carcinoma (mRCC). Early C‐reactive protein (CRP) kinetics, especially the recently introduced CRP flare‐response phenomenon, has shown promising results to predict IO efficacy in mRCC, but has only been studied in second line or later. Here, we aimed to validate the predictive value of early CRP kinetics for 1st‐line treatment of mRCC with αPD‐1 plus either αCTLA‐4 (IO+IO) or tyrosine kinase inhibitor (IO+TKI). Methods In this multicentre retrospective study, we investigated the predictive potential of early CRP kinetics during 1st‐line IO therapy. Ninety‐five patients with mRCC from six tertiary referral centres with either IO+IO (N = 59) or IO+TKI (N = 36) were included. Patients were classified as CRP flare‐responders, CRP responders or non‐CRP responders as previously described, and their oncological outcome was compared. Results Our data validate the predictive potential of early CRP kinetics in 1st‐line immunotherapy in mRCC. CRP responders, especially CRP flare‐responders, had significantly prolonged progression‐free survival (PFS) compared with non‐CRP responders (median PFS: CRP flare‐responder: 19.2 months vs. responders: 16.2 vs. non‐CRP responders: 5.6, P < 0.001). In both the IO+IO and IO+TKI subgroups, early CRP kinetics remained significantly associated with improved PFS. CRP flare‐response was also associated with long‐term response ≥ 12 months. Conclusions Early CRP kinetics appears to be a low‐cost and easy‐to‐implement on‐treatment biomarker to predict response to 1st‐line IO combination therapy. It has potential to optimise therapy monitoring and might represent a new standard of care biomarker for immunotherapy in mRCC.


INTRODUCTION
First-line treatment of metastatic renal cell carcinoma (mRCC) has changed substantially in recent years because of the introduction of a new therapy regimen, mainly based on immune checkpoint inhibition (IO). [1][2][3] Currently, two different types of approved first-line combination therapies are applied equivalently for the treatment of intermediate and poor-risk metastatic mRCC according to IMDC (International Metastatic Renal Cell Carcinoma Database Consortium Score): (1) a combination of aPD-1 and aCTLA-4 immune checkpoint inhibitors as well as (2) a combination of aPD-1 (or aPD-L1) with small-molecule tyrosine kinase inhibitors (TKI) targeting the vascular endothelial growth factor receptor (VEGFR). [4][5][6][7][8] In essence, these two regimens can be classified as an intensified immune checkpoint inhibition (IO+IO) and a combination of immune checkpoint inhibition plus anti-angiogenic therapy (IO+TKI).
However, only a subset of patients responds to these first-line IO combination therapies. On the one hand, reliable predictive biomarkers could identify early therapy failure, which is of high clinical relevance. On the other hand, severe unnecessary side effects could be avoided, and the individual therapy regimen could be further optimised.
In general, IO treatment success is based on the induction of an antitumor immune response. Creactive protein (CRP) is a serum acute-phase reactant and clinically widely used surrogate biomarker for the assessment of systemic inflammation. The occurrence and kinetics of systemic inflammatory response reflected by serum CRP has been implicated with clinical outcome and treatment response in diverse cancer entities, including urothelial cancer, non-small-cell lung cancer and mRCC. [9][10][11][12][13][14] Several studies investigated CRP levels at initial diagnosis or baseline before therapy initiation and associated increased systemic inflammation with poor oncologic prognosis. As cancers can also induce chronic inflammation, on-treatment CRP kinetics may have predictive value for immunotherapy treatment success. 15,16 Just recently, Fukuda et al. described the CRP 'flare-response' phenomenon defined by an early CRP increase after IO treatment initiation with a subsequent drop below baseline. These early CRP changes appear to mirror the dynamic phase of systemic inflammation after inducing the desired antitumoral immune response on IO therapy. 15 Of note, this novel concept allowed an accurate prediction of therapy success in 42 mRCC patients treated with aPD-1. However, the investigated cohort only included a limited patient number and aPD-1 monotherapy was administered as 2nd-line (or later) post-TKI treatment. As IO monotherapy will occur less frequently in the future, our study aimed to investigate the emerging phenomenon of CRP flare-response in a multicentre mRCC cohort receiving either IO+IO or IO+TKI as 1st-line standard of care therapy.

Survival analysis by early CRP kinetics
The median progression-free survival (PFS) after initiation of IO treatment was 5.6 months (95% CI 3.4-12.2 months) for non-CRP responders, 16.2 months for CRP responders (95% CI 10.9 monthsnot reached) and 19.2 months for CRP flare-responders (95% CI 17.4 monthsnot reached) and differed significantly (Figure 2b). When the overall cohort was divided into subgroups of patients receiving IO+IO or IO+TKI, early CRP kinetics remained significantly associated with PFS on immunotherapy in both groups (Supplementary figure 1). Of note, the majority of CRP flare-responders (8/10 = 80.0%) showed long-term therapy response lasting ≥ 12 months. Thus, the mean duration of IO response differed significantly between the groups (P = 0.001; Figure 3a).

Cox regression
In the univariate Cox regression, early CRP dynamics was the only factor that was significantly associated with the PFS, besides ECOG score (Table 2). Of note, baseline CRP level was not associated with PFS, but highest in the CRP response group (Figure 3b, P < 0.001). Compared to Non-CRP responders, CRP responders had a risk reduction for progression of 68% [hazard ratio HR 0.32, 95% confidence interval (CI) 0.17-0.62, P = 0.001] and CRP flare-responders of 73% (HR 0.27 95% CI 0.11-0.66, P = 0.004). No other patient or tumor-related factor had an impact on the PFS after IO treatment initiation. In the multivariate Cox regression model, the impact of CRP dynamics and ECOG score remained significant (Table 2). Additionally, the therapy regimen, baseline CRP (HR 1.01, P = 0.021) and T stadium (T2 vs. T1: OR 7.56, P = 0.004) now also had a significant impact on progression.
Regarding OS, only ECOG had a significant impact in the univariate Cox regression, as patients with worse performance status had an increased risk for death from any cause (Supplementary table 1). However, this association did not remain significant in the multivariate Cox regression model.

DISCUSSION
In this retrospective multicentre study, we validate that early CRP kinetics on immunotherapy is a promising predictive biomarker in mRCC. Because of its low cost and wide clinical availability, the CRP kinetic assessment is easy to implement into daily clinical practice and may prove to be a valuable tool for IO therapy monitoring in the future.
In our cohort consisting of 95 patients with either IO+IO-or IO+TKI-based first-line therapy, CRP flare-response was associated with long-term response and improved PFS in the aPD-1-based first-line setting of mRCC. However, in our mRCC cohort, early CRP kinetics showed no significant association with OS, which is most likely attributed to the relatively low number of events in the cohort. Since the new 1st-line combination therapies in mRCC remarkably prolong OS, we plan to reanalyse this cohort after extending the follow-up period. Further, early CRP kinetics was significantly associated with improved PFS in both subgroups (IO+IO and IO+TKI), leading us to conclude that early CRP kinetics is a robust predictive biomarker in mRCC independent of the chosen first-line treatment combination. Since Fukuda et al. described the predictive value of early CRP kinetics for nivolumab monotherapy in 2nd line or later, it appears that early CRP kinetics can therefore be used to optimise treatment monitoring for all aPD-1-based therapies in mRCC. 15 We consider this to be a particularly important information for the daily clinical routine, as early CRP kinetics could be used as a simple and cost-effective biomarker for all immunotherapy regimen in mRCC. Non-CRP response would lead to earlier staging, and in the event of tumor progression, allow clinicians to administer alternative and more effective therapies while preventing exposure to potentially life-threatening toxic effects of immunotherapy. 17 In our analysis, early CRP kinetics appear to have the potential to predict treatment response before initial staging and thus lead to earlier treatment modification, which could ultimately improve the clinical course of mRCC patients.
In addition, it appears to be highly relevant to sensitise clinicians to the characteristic CRP flareresponse phenomenon, as a rapid increase in CRP could be the result of a desirable antitumor immune response. CRP flare-response should, in the absence of other clinical symptoms, thus not be misinterpreted as a bacterial infection or another side effect after IO therapy initiation especially since antibiotic-induced dysbiosis can compromise the clinical activity of immunotherapy by modulating, for example the gut microbiome. 18 Exploring the tumor immunologic basis of the differential CRP kinetics after initiation of immunotherapy might further enhance our understanding of the interplay between the RCC tumor cells and its tumor microenvironment (TME). [19][20][21] Baseline serum CRP concentration, which may reflect the baseline RCC immunogenicity, differs significantly between the CRP response groups. The low baseline CRP level in flare-responders could be an indirect surrogate for low or absent chronic inflammation caused by the tumor burden. Thus, we hypothesise that in treatment-na€ ıve RCC tissue, differential immune phenotypes may predict early CRP kinetics as IO treatment triggers distinct immune cell infiltration patterns to enrich the TME. Thereafter, the induction of an antitumor immune response leads to systemic inflammation through the release of inflammatory mediators, which can ultimately be measured by serum CRP. To address this hypothesis, future studies will need to perform comprehensive phenotyping of treatment-na€ ıve tumor tissue, followed by integration of the early CRP kinetic subset. From a clinical point of view, the identification of specific TME patterns in treatment-na€ ıve RCC tissue that robustly predict early CRP kinetics and response would be of high relevance to stratify our patients before therapy, especially since currently available predictive tools such as PD(L)-1 immunohistochemistry (IHC) only play a minor role in mRCC. 22 From a cancerimmunologic point of view, it would be of high relevance to identify the distinct immune signatures associated with non-CRP response and IO treatment failure to identify potential targets for tailored combination therapy in this immunotherapy-unresponsive RCC subgroup. Increased baseline concentration of inflammation markers such as CRP or IL-8 before oncological treatment has also been associated with worse clinical outcome in mRCC patients treated with immunotherapy elsewhere, but the dynamic and early change in systematic inflammation after therapeutic intervention was mostly neglected. 9,10,12,23,24 In mRCC, an early decrease in CRP after initiation of TKI therapy has already been associated with improved response and survival. 25 Only recently, the predictive potential of characteristic longitudinal changes in CRP, especially the newly described flare-response, during the first 3 months of aPD-1 monotherapy in the post-TKI setting has been highlighted. We evaluated the predictive value of early CRP dynamics in a larger, multicentre and more clinically relevant cohort in the first-line setting in mRCC and demonstrated that CRP responders and particularly CRP flare-responders showed favorable progression-free survival (PFS) and mostly durable treatment response. Further studies will have to clarify whether the flareresponse kinetics of systemic inflammation can be sharpened by replacing the relatively nonspecific CRP with other acute-phase reactants or immune mediators. After prospective validation of the predictive potential of early CRP kinetics in mRCC and possibly in additional tumor subtypes, we propose early CRP kinetics as a promising ontreatment biomarker for stratifying our patients in the era of immuno-oncology.
Despite noteworthy strengths, such as the multicentre approach and the comparably large study cohort, our study also has several limitations. First and foremost, we acknowledge that the study is limited by its observational nature and the relatively short follow-up time, especially for the meaningful endpoint OS. Moreover, our results should be interpreted within the limitations of the retrospective design. CRP was measured in different routine clinical laboratories at the study centres and without a standardised scheme, so some CRP flare-responses may have been missed. In addition, modification of the new and not prospectively validated early CRP kinetic concept might increase its predictive value. Nevertheless, we propose a prospective evaluation of our results in future studies, based on our promising retrospective data.
If prospectively validated, we propagate that early CRP kinetics should be assessed as an easyto-implement, non-invasive biomarker during IO combination therapy in mRCC as the new standard of care, as early detection of treatment success and failure might have the potential to optimise treatment monitoring and adjustment and to prevent exposure to potentially lifethreatening side effects of IO therapy.

METHODS
In this retrospective multicentre study, N = 118 consecutive mRCC patients from six German tertiary referral centres receiving either first-line IO+IO (aPD-1/nivolumab + aCTLA4/ ipilimumab) or IO+TKI (a PD-1/pembrolizumab + VEGFR-TKI/ axitinib) were screened. Patients with CRP measurements at baseline (closest to treatment initiation, maximum 6 weeks before), at least once within the first month of treatment and at least one further CRP at the time of first staging or clinical progression were included in the study. Of the total N = 118 patients initially studied, N = 23 were excluded due to missing CRP values, resulting in a study cohort of N = 95 patients. This study was conducted according to the Declaration of Helsinki and approved by the responsible ethical review board (reference #20201211-01).
The patient demographics and baseline parameters including IMDC risk criteria were obtained. Tumor response was graded according to response evaluation criteria in solid tumors (RECIST v1.1). 26 Therapy outcomes were compared among the three characteristic therapy groups, defined by diverging CRP dynamics. According to the earlier definition by Fukuda et al., 'CRP flare-responders' were defined as an early increase in CRP levels to more than double from baseline within 1 month after therapy initiation and a subsequent decrease below the baseline within 3 months. Patients with a decrease by ≥ 30% from baseline within 3 months without flare-response were classified as 'CRP responders', all other patients as 'non-CRP responders' (Figure 1). 15 To define these CRP dynamic groups, CRP at baseline, during the first month after treatment initiation and follow-up visits was obtained. Serum CRP concentration was measured in accredited routine laboratories in each participating centre and is given in mg dL À1 (clinical reference < 0.5 mg dL À1 ).
Categorical variables were reported as frequencies and proportions, continuous data as the median and range. Fisher's exact tests, Mann-Whitney U-tests and Kruskal-Wallis tests were applied to perform intergroup comparisons. The PFS and OS, including 95% confidence intervals, were estimated from the day of treatment initiation until the respective event using the Kaplan-Meier method and compared with log-rank tests. Progression was defined according to the RECIST v1.1 criteria including death from any cause. To compare the impact of the therapy regimen (IO+IO vs. IO+TKI), CRP dynamics (CRP flare-responder, CRP responder vs. non-CRP responder), baseline patient (age, gender, ECOG) and tumor-related parameters (e.g. IMDC, histology, pT-stage) on OS and PFS, univariate and multiple Cox regressions were conducted. Patient age and CRP baseline were defined as continuous, all others as categorical variables. In the event of missing data, cases were excluded from the analysis. Statistical analyses were performed with SPSS version 25 (IBM, Armonk, NY, USA), R (version x64 4.0.3) and GraphPad Prism 9 (GraphPad Software Inc, CA, USA). All statistical tests were two-sided, and P-values < 0.05 were considered significant.