Advances in immunotherapy have transformed practice in relapsed and/or metastatic (R/M) head and neck squamous cell carcinoma (HNSCC). Two anti-PD1 antibodies, nivolumab and pembrolizumab, have been shown to improve overall survival (OS) in patients with R/M HNSCC and have been approved for use in the second-line setting based on the results from the CheckMate-141 and KEYNOTE-040 trials, respectively.1 2 Recently, given the observed efficacy in the KEYNOTE-048 trial, pembrolizumab in combination with platinum and 5-fluorouracil has been approved by the US Food and Drug Administration and European Medicines Agency for first-line treatment in R/M HNSCC and as monotherapy for PD-L1-positive population.3 Other trials of immune checkpoint inhibitors (ICIs) or combinations with other immune-stimulatory drugs are also tracking first-line and second-line benefit.4 Although ICIs present a promising new opportunity, objective responses remain around 20% in first-line and second-line patients with R/M HNSCC unselected for tumor PD-L1 expression. In addition, despite data showing that PD-L1 correlates with better efficacy of ICIs, no other reliable biomarker for effectively selecting optimally responsive patients has been identified as yet. Moreover, no on-treatment monitoring biomarker has yet been defined.

With the aim of selecting eligible patients with advanced cancer for early clinical trials, different tools have been developed. The Royal Marsden Hospital (RMH) score established a prognostic score based on three objective baseline markers: lactate dehydrogenase (LDH) level, number of metastatic deposits (≤2 vs >2) and albumin level.5 6 This prognostic score was able to stratify patients into two distinct groups according to their risk of death. More recently, the Institut Gustave Roussy’s Early Drug Development Department developed a new prognostic immune score (Gustave Roussy Immune Score -GRIm-Score-) applied to ICI phase I trial patients. This score used LDH, albumin and neutrophil-to-lymphocyte ratio (NLR).7 Both scoring systems have been retrospectively and prospectively validated in different cohorts but remain globally underused by clinicians and investigators.8 9 Other measures have been proposed in the past, including the MD Anderson experience, which included low hemoglobin level as an independent factor of poor survival,10 and the Princess Margaret Hospital Index, which used Eastern Cooperative Oncology Group (ECOG) performance status.11

There is increasing interest in NLR as a potential predictive marker in patients treated with ICI. Many studies have identified high NLR as a poor prognostic factor in several malignancies, and a large meta-analysis has shown its prognostic impact on progression-free survival (PFS) and OS in different disease subgroups, including HNSCC, independent of tumor site and stage.12 13 However, the most discriminatory cut-off point for NLR remains unknown and may differ between different tumor sites and stages.

Fibrinogen is a 340 kDa glycoprotein that is primarily produced by hepatic cells and functions as an important clotting factor in the haemostatic pathway. Fibrinogen is converted into fibrin through the proteolytic effect of thrombin.14 It also serves as an acute-phase protein which rises in response to systemic inflammation and tissue injury.15 Recent studies have suggested that elevated fibrinogen promotes cancer cell growth, progression and metastasis.16–22 Furthermore, fibrinogen levels have been associated with tumor size and lymph node spread in various cancer types, including HNSCC.23 24 Fibrinogen biosynthesis is regulated by constitutive, hormonal and cytokine-mediated mechanisms. As an acute-phase protein, fibrinogen biosynthesis is upregulated by interleukin (IL)-6.25 An emerging role of IL-6 in modulating multiple functions of immune cells including T cells, dendritic cells, and macrophages is thought to contribute to dysfunctional innate and adaptive immunity against cancer cells.26 Indeed, elevated levels of IL-6 have been observed in patients with various types of cancer and have been considered a poor prognostic factor, given its detrimental effects on antitumor immunity.27 Other cytokines, such as tumor necrosis factor alpha (TNF-α) and IL-1, suppress fibrinogen synthesis.28 TNF-α and IL-1 contribute to antitumor activity by promoting tumour-associated macrophage differentiation to M1 phenotype, by activating T effector cells and by attracting neutrophils and monocytes to the tumor site.29

We conducted a study to evaluate the usefulness of pretreatment and on-treatment serum fibrinogen levels alongside previously described prognostic markers (NLR and LDH) to establish an accurate scoring system predictive of response and survival in patients with R/M HNSCC treated with ICI.

Our study successfully identified inflammatory markers which are predictive of response and survival in patients with R/M HNSCC treated with ICI. These parameters could potentially be used in the clinic to improve patient monitoring. At pretreatment time point, fibrinogen and NLR showed to be relevant markers of poor OS. However, the selected NLR cut-off point (NLR≥8) was higher than that used in the GRIm-Score (NLR≥6).7 This may be explained by the fact that patients with HNSCC, especially those with local and/or regional relapse, frequently present with ulcerative and overinfected lesions that lead to an increased neutrophil count.31 NLR has been described as a prognostic factor in multiple solid cancers. The meta-analysis conducted by Templeton et al12 confirmed this among a wide range of malignancies, although on a different magnitude for each subsite. In addition, NLR has also been shown to be a predictive marker in patients treated with ICI. Studies of metastatic melanoma suggest that baseline NLR can predict response to ipilimumab and nivolumab.32 33 Takenaka et al¹³ quantified the prognostic value of NLR in patients with HNSCC from a pooled analysis of 16 studies and showed that higher NLR was associated with worse outcome (pooled HR ranging from 1.47 to 1.93). However, the population comprised patients with early or locally advanced HNSCC, and the cut-off values of NLR for dichotomization ranged from 1.92 to 5.0.

In order to translate a continuous variable such as NLR or serum fibrinogen into a clinical decision tool, it is necessary to determine a cut-off point to stratify patients into distinct groups.34 Although the mean or median value of the marker has been frequently used before, it has been deemed inappropriate, and it is often desirable to determine cut-off points by optimizing the correlation with clinical outcomes by using minimization of p values or maximization of sensitivity and specificity tests.35 Regrettably, among the studies of the prognostic impact of NLR, including publications such as the GRIm-Score,7 the methods for cut-off determination remain obscure or unmentioned in most instances. Conversely, in our study, we implemented a cut-point estimation method that can be easily accessible and reproducible using the ‘maxstat’ package available at CRAN (https://CRAN.R-project.org/package=maxstat).

Our multivariate analysis results show that on-treatment fibrinogen and NLR are the most reliable biomarkers to predict both disease progression and mortality in our population. This may reflect the fact that early changes in the immune system following initiation of immunotherapy, rather than pretreatment characteristics, give better representation of the underlying biological process.36 This is in line with the results of the recently published melanoma study by Li et al,30 which showed that the change in the NLR during treatment with ICI is associated with survival30 in a non-linear fashion, with moderate reductions of NLR correlated with longest OS. Moreover, our data showed that lower levels of on-treatment NLR and fibrinogen levels correlate with response. An on-treatment NLR cut-off point of 3 enables the stratification of patients according to their response to ICI with a sensitivity of 0.9. On the other hand, despite being unable to find a clinically significant on-treatment fibrinogen cut-off for response, this is the first study to demonstrate a relationship between fibrinogen levels and response during ICI treatment.

To our knowledge, this is the first published study to evaluate the prognostic value of serum fibrinogen levels in patients with R/M HNSCC treated with ICI. IL-6 signaling is a critical mechanism for the induction of dysfunctional immune responses.27 IL-6-mediated STAT3 activation has been reported to regulate the expansion of myeloid-derived suppressor cells (MDSCs) and the expression of PD-L1 in cancer and immune cells.30 Considering the established stimulatory action of IL-6 on fibrinogen production,25 measuring serum fibrinogen could predict the levels of IL-6 and circulating MDSCs in patients, and it may constitute a method for evaluating the state of host systemic immune response. Our ‘on-treatment immune score’ showed that using on-treatment fibrinogen improved the discriminatory power when it is taken into account together with NLR. Importantly, NLR and serum fibrinogen are cheaper and faster laboratory measurements than many other biomarkers, and they can be readily applied in daily clinical practice.

One might argue that baseline prognostic scores would normally be preferred to on-treatment score models. However, in a day-to-day clinical scenario, this may not always be the case, even less so when applied to patients with R/M HNSCC. According to the KEYNOTE-048 results,3 the decision-making for first-line ICI treatment options should take into account PD-L1 Combined Positivity Score (CPS), alongside other important clinical considerations such as performance status, disease burden and progression rate.37 However, even in patients with high CPS (≥20), the response rate remains relatively low (23%) and careful review of the Kaplan-Meier curves from this study shows that a considerable proportion of PD-L1-positive (CPS≥1) patients treated with pembrolizumab monotherapy progressed or died earlier than those treated with chemotherapy. Therefore, an early identification of those patients that present poor prognostic features within the first month of treatment could assist the clinician in making informed decisions regarding risk-adaptive treatment changes and could add valuable information in the scenario of pseudo-progression.

We acknowledge that this study is limited due to its retrospective nature, its relatively small numbers, and the lack of data on PD-L1 status. Nevertheless, the statistical analysis is powerful enough and the scoring system exhibits an appropriate discriminative ability. Moreover, internal cross-validation and model evaluation confirmed good accuracy and concordance of the score. All the confounding variables, including those that could affect the fibrinogen measurement accuracy (smoking status, liver dysfunction and anticoagulation treatment), have been controlled and the impact of fibrinogen on survival was confirmed in a sensitivity analysis according to p16 status. In addition, this is the first scoring system designed to be used in the setting of patients with R/M HNSCC and the first that has described the prognostic value of serum fibrinogen as a surrogate marker of the state of the host immune system. We acknowledge that, despite showing appropriate internal cross-validation of our model, the scoring system should now be validated using independent prospective datasets. During the conduct of this research, we approached two pharmaceutical companies who had conducted phase II and III clinical trials of ICI in the first-line and second-line treatment of R/M HNSCC and requested access to data in order to validate our model. Regrettably, these attempts were unsuccessful, in part because of a prevailing view that the current regulatory landscape precludes sharing of data for independent validation. Therefore, we encourage the research community to test the performance of our model in their independent datasets.

This is the first study to demonstrate the relationship between serum fibrinogen and survival during ICI treatment for patients with R/M HNSCC. We show that a scoring system using on-treatment NLR for OS and PFS improved its discriminatory power when on-treatment fibrinogen is taken into account. Because NLR and serum fibrinogen are obtained from routine CBC, their use does not involve additional procedures or extra costs for healthcare providers. Implementation of the on-treatment score in clinical practice could allow better monitoring and optimal clinical management of patients with R/M HNSCC receiving first-line or second-line systemic treatment with ICI. Prospective studies with a larger patient cohort are needed for validation.