Systemic inflammatory biomarkers as prognostic tools in patients with gastroesophageal adenocarcinoma

Purpose Gastroesophageal adenocarcinoma is associated with poor prognosis, even in resectable stages. Systemic inflammation plays a key role in cancer progression. Yet, information on prognostic values of systemic inflammatory parameters in European cohorts is scarce. Methods We analysed systemic inflammatory biomarkers (neutrophil-to-lymphocyte ratio (NLR), leucocyte-to-lymphocyte ratio (LLR), platelet-to-lymphocyte ratio (PLR), systemic inflammation response index (SIRI) and modified Glasgow Prognostic Score (mGPS)) at the time of cancer diagnosis and their association with overall survival (OS) in patients with gastroesophageal adenocarcinoma treated at the Medical University of Vienna between 1990 and 2020. Results In this analysis of 769 patients with gastroesophageal adenocarcinoma, higher mGPS (0–2) scores were associated with shorter OS in the overall cohort (24.9 versus 11.9 versus 7.6 months; HR 1.74, 95% CI 1.549–1.056; p < 0.001), in locally advanced (31.1 versus 19.8 versus 13.9 months, HR 1.561, 95% CI 1.274–1.912; p < 0.001) and in advanced/metastatic settings (12.3 versus 7.3 versus 5.8 months; HR 1.377, 95% CI 1.777–1.611; p < 0.001). In multivariate analyses, the association of mGPS with the OS stayed statistically significant in the locally advanced cohort (HR 1.397, 95% CI 1.068–1.828; p = 0.015), whereas NLR, LLR, PLR and SIRI did not. mGPS was associated with more advanced stages (p < 0.001) and weight loss (p = 0.002). Conclusion mGPS poses a feasible prognostic tool in patients with locally advanced gastroesophageal cancer. Supplementary Information The online version contains supplementary material available at 10.1007/s00432-023-05424-4.


Introduction
Gastroesophageal cancer represents a major global disease burden with high mortality rates (Sung et al. 2021).Although recent advances in treatment strategies have improved overall survival, prognosis remains poor, even in resectable settings (Lordick et al. 2022;Obermannová et al. 2022).Thus, novel prognostic markers are warranted to improve patient management (Dhakras et al. 2020).
The rise of checkpoint inhibition did not only lead to several practices-changing drug approvals, but also raised awareness of local as well as systemic inflammatory parameters as predictive and prognostic parameters in several oncological entities including gastroesophageal cancers (Moehler et al. 2021;Shitara et al. 2022;Li et al. 2021).Systemic inflammatory biomarkers, including leucocyte ratios, such as neutrophil-lymphocyte ratio (NLR), lymphocyte-leucocyte ratio (LLR), platelet-lymphocyte ratio (PLR) and the systemic inflammation response index (SIRI; neutrophils*monocytes/lymphocytes) as well as the C-reactive protein (CRP) and albumin-based modified Glasgow prognostic score (mGPS), propose potential association with treatment outcome (Laird et al. 2013;Proctor et al. 2011).Retrospective analyses in Asian cohorts showed that systemic inflammation might be an important prognostic factor independent of tumour stage and treatment regimen (Matsumoto et al. 2023;Liu et al. 2021;Huang et al. 2020).However, large real-world investigations of Western cohorts comprising all tumour stages are lacking as existing data focus largely on prediction of surgical as well as pathological outcomes and do not investigate inflammatory markers in association with other known prognostic factors in multivariate settings (Kim et al. 2020;Schietroma et al. 2022;Nechita et al. 2022;Pikuła et al. 2022;Jomrich et al. 2021;Petrillo et al. 2018).As tumour biology, biomarker expression, treatment response and survival differ between Asian and Western cohorts (Rha et al. 2022;Zhang et al. 2022b), the aim of this analysis was to investigate systemic inflammatory processes at the time of first cancer diagnosis as prognostic markers in a large European cohort of gastroesophageal cancer patients.

Patient recruitment
Adult patients with gastroesophageal adenocarcinoma diagnosed and/or treated at the Department of Medicine I -Division of Oncology at the Medical University of Vienna between 01.01.1990 and 31.12.2020 with available laboratory parameters were included in this analysis independent of stage and tumour location.In order to reduce selection bias and improve representability, patients with isolated missing values were not excluded from this analysis.Clinical data including laboratory parameters were retrieved by structured electronical chart review.Disease was pathologically confirmed by expert pathologists.Analyses of human epidermal growth factor receptor 2 (HER2) and combined positive score (CPS) for PD-L1 testing have been routinely performed in metastatic settings since 2012 and 2018, respectively.Helicobacter pylori screening is not routinely performed; however, when infection is evident in microscopic specimen, it is reported.
Treatment strategies were reviewed by an interdisciplinary tumour board and patients received the best available treatment according to current treatment guidelines at the time of diagnosis.
Death was registered according to hospital chart data and/ or public data provided by Statistik Austria.Patients without registered date of death at the time of data cut-off were censored at the date that they were last known to be alive.Overall survival (OS) was defined as the time from initial cancer diagnosis to patient's death or last follow-up date.
In addition to analysis of the overall cohort, subgroups based on staging subgroups were performed to further investigate prognostic impact of systemic inflammatory parameters.Patients were divided into localised (stage 1), locally advanced (stage 2 and stage 3) and advanced/metastatic (stage 4) subgroups based on the latest (8th) edition of UICC TNM staging classification.This ensured minimization of potential bias due to improved diagnostic procedures over time and is in line with application of treatment strategies.Patient data were collected at the time of first diagnosis of gastroesophageal cancer, even if the patient experienced cancer recurrence throughout their history.Patients were only recorded once, also when they received multiple gastroesophageal tumours throughout their history.-All procedures were performed in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1964 and latter versions.The study was approved by the ethics committee of the Medical University of Vienna (reference number: 1600/2021).
NLR was defined as simple ratio between absolute neutrophil and absolute lymphocyte counts, LLR as simple ratio between absolute lymphocyte count and absolute leucocyte count and PLR as simple ratio between absolute platelet count and absolute lymphocyte.SIRI was defined as absolute neutrophil count multiplied by absolute monocyte count divided by absolute lymphocyte count.mGPS was scored 0 when CRP was < 1.0 mg/dl, 1 when CRP was > 1.0 mg/dl and albumin was ≥ 35 g/L, and scored 2 when CRP was > 1.0 mg/dl and albumin was < 35 g/L.For further evaluation, patients were divided into high and low groups according to median levels of NLR, LLR, PLR and SIRI (≤ median or > median) respective to the subgroup.
Blood taken for the analysis was performed regularly before administration of therapy and/or at diagnosis and evaluated retrospectively with structured electronical chart review.Laboratory parameters were classified as being conducted at first diagnosis, when the blood draw took place at the time of histopathological first cancer diagnosis or within 1 month before or after this date.When blood was taken within 1 month before or 1 month after histopathological first cancer diagnosis date, it was only considered for this analysis when no tumour-specific therapy (i.e.surgery, systemic therapy, radiation therapy) had taken place in this time frame.When multiple laboratory analyses were available, the one which was closest to histopathological first diagnosis was considered for this analysis.
As more severe disease stages as well as tumour location, sarcopenia, helicobacter infection, nicotine and alcohol abuse have been associated with higher systemic inflammatory parameters, associations between NLR, LLR, PLR and mGPS with cancer staging subgroups (localised, locally advanced, advanced or metastatic) as well as weight loss and BMI were performed.

Statistical analysis
Upon retrospective data collection, results were entered into the statistical package for the social sciences (SPSS) 20.0 software (SPSS Inc., Chicago, IL, USA).
Descriptive analyses of clinical data were performed by mean (SD) for metric and median for nominal data.Calculations for survival differences were performed with log-rank test for nominal variables and with Cox regression for metric variables.Visualisation of survival analyses was performed with Kaplan-Meier product.Cox proportional hazard model was used for multivariate analysis, which included all variables with statistically significant results in univariate analysis.For further association analyses between staging subgroups, weight loss and BMI with systemic inflammatory ratios, the chi-square test and Kruskal-Wallis test were performed as appropriate.P-values < 0.05 were considered to indicate statistical significance.Due to the hypothesis generating design of the studies, no correction for multiple testing was applied unless specified otherwise (Bender and Lange 2001).

Patient and tumour characteristics
In this analysis, we included 769 patients with gastroesophageal adenocarcinoma.68% of patients were male.The mean age was 63.9 years (SD 12.2).49% of tumours were located in the stomach, 37% in the gastroesophageal junction area and 14% in the oesophagus.Concerning staging, 101 (13%) patients had a localised disease (UICC stage 1), 361 (47%) patients were categorised as locally advanced disease (stage 2: n = 125, stage 3: n = 236) and 307 (40%) patients had a primary advanced or metastatic disease (UICC stage 4) at initial cancer diagnosis.Patient and tumour characteristics for localised, locally advanced and advanced or metastatic subgroups are available in Supplementary tables 1 + 2.
Further patient, tumour characteristics and their association with the OS in the overall cohort as well as in subgroups can be seen in Supplementary Tables 1 and 2.

Systemic inflammation
Laboratory parameters for the overall cohort and subgroups are available in Supplementary Table 3.
Neutrophil-lymphocyte ratio (NLR) was available in 598 patients, lymphocyte-leucocyte ratio (LLR) in 594 patients, platelet-lymphocyte ratio (PLR) in 567 patients and systemic inflammation response index (SIRI) in 597 patients.Mean and median values for inflammatory ratios are available in Table 1.
More severe disease stages were associated with higher rates of weight loss at initial diagnosis (39% versus 51% versus 61%; p < 0.001), but not with BMI (p = 0.108).However, lower BMI levels were statistically significantly associated with weight loss at first diagnosis (p < 0.001).
In addition, NLR, LLR, PLR and SIRI were grouped into high level (> median) and low level (≤ median) cohorts based on the medians of the respective subgroups (see Table 1).There was a statistically significant association between higher levels and OS in the overall as well as locally advanced and advanced or metastatic subgroups (see Tables 1 and 2).Kaplan-Meier survival curves are shown in Supplementary Figs.6-9.
For resectable setting, no multivariate analysis was performed as no variables had statistically significant association with OS in univariate analyses.

Discussion
Inflammatory parameters play a key role in cancer development and progression.Yet, prognostic effects of systemic inflammatory parameters have rarely been studied in large Western cohorts with gastroesophageal adenocarcinoma and it is unclear which inflammatory parameter is the best predictor for overall survival.The results of our analyses shed some light on this issue, as they show that mGPS is a feasible prognostic tool to estimate overall survival in a large, representable European cohort of gastroesophageal adenocarcinoma patients.
Although the leucocyte ratios NLR, LLR and PLR also showed prognostic potential in univariate analyses, only high mGPS scores at first diagnosis could be identified as an independent tool in multivariate analysis of the locally advanced cohort.NLR and LLR are surmised to reflect an increased inflammatory response to tumour cells (Buonacera et al. 2022).Whether this mechanism is associated with immune cell infiltration in tumour tissue is widely discussed (Choi et al. 2016;Puhr et al. 2022), but no clear association has been proven yet.In addition, PLR may represent the important role of platelet-derived tumour angiogenesis maintained by vascular endothelial growth factor and platelet-derived growth factor leading to tumour progression and dissemination.Thus, several trials have shown distinctive associations of these ratios with tumour progression and outcome in gastroesophageal cancer patients as well as patients with other cancer entities (Namikawa et al. 2016;Starzer et al. 2021;Pikuła et al. 2022;Miyamoto et al. 2018;Lian et al. 2015;Kim et al. 2022;Petrillo et al. 2018).Yet, the definition of clear cut-off levels for risk stratification is still lacking (Schiefer et al. 2022).This might be due to the fact that although higher levels are associated with more severe tumour stages and survival, differences between ratios are small and effected by outliers (see Supplementary Figs.1-3).Systemic inflammation parameters not only serve as prognostic tools but could also be used to predict response to immunotherapy.Initial emerging data from translational research programmes of clinical trials indicate potential usability of these markers, particularly NLR, in decisionmaking between chemotherapy and immunotherapy doublet (Chau et al. 2022).Although the implementation of these ratios outside of academic evaluations is still not feasible, further translational research of clinical trials will provide valuable data and an improved understanding of these markers for the clinical routine.
The mGPS, which is based on CRP and albumin levels, presents a more robust and feasible tool than systemic inflammatory ratios as it is easily calculated and only encompasses three levels (score 0 to 2).mGPS has been shown to be associated with postoperative outcome in gastroesophageal cancer patients in several Asian cohorts (Jiang et al. 2012;Chen et al. 2020;Hirashima et al. 2014;Zhang et al. 2022a, b).Yet, data on Western cohorts and comparisons between tumour stages are scarce.
In our cohort, mGPS was also associated with more severe tumour staging.Further analyses of stage subgroups showed that mGPS was an independent prognostic marker in multivariate analysis of the locally advanced cohort.These results suggest that although scores are the highest in advanced and metastatic settings, the prognostic value is the greatest in locally advanced settings.This finding is in line with perioperative studies of Asian cohorts (Jiang et al. 2012;Chen et al. 2020;Hirashima et al. 2014;Zhang et al. 2022a, b).Recently published data from a Romanian cohort, investigating leucocyte ratios, also suggests that the prognostic value of inflammatory biomarkers is the greatest in locally advanced settings (Nechita et al. 2022).This is further supported by two European studies investigating inflammatory biomarkers in resectable gastric cancer (Pikuła et al. 2022;Schietroma et al. 2022).However, multivariate analyses and recommendations for clear cut-off levels of leucocyte ratios are lacking in these trials.Furthermore, the role of tumour location in systemic inflammatory processes is unclear.
Thus, based on our findings, mGPS might pose a feasible prognostic tool in everyday practice and could improve patient management in perioperative settings by raising physicians' awareness of early recurrences as well as perioperative morbidity in patients with higher scores.Yet, the exact mechanisms behind these consistent results of associations between elevated systemic inflammatory biomarkers and inferior outcome in locally advanced settings have not been identified and need to be further explored.
In a recently published analysis of advanced gastroesophageal cancer patients undergoing first-line treatment from the phase III EXPAND trial, mGPS could be associated with sarcopenia (Hacker et al. 2022).Our results show that elevated systemic inflammation parameters were associated with weight loss rather than with BMI.Although not directly assessed in the current cohort, sarcopenia, which is often characterised by weight loss independent of BMI, might also be a large contributor to higher mGPS scores in our cohort.However, concerning multivariate analyses in the overall cohort, mGPS stayed statistically significantly associated with the OS and so did weight loss.Thus, the prognostic value of mGPS might not only be associated with sarcopenia but other values.This theory is supported by the findings of Hacker et al. as mGPS carried full prognostic information for OS, but direct causal path from sarcopenia to survival was lacking (Hacker et al. 2022).
Moreover, elevated systemic inflammatory parameters and consequently higher mGPS levels might be caused by circulating tumour cells in more advanced stages, as these tumours are known to be associated with higher levels of circulating tumour components.Circulating tumour DNA (ctDNA) has gained scientific interest as a key prognostic value after tumour resection in various cancers (Moati et al. 2021).First attempts to guide adjuvant therapy by the analysis of ctDNA in colorectal cancer have shown promising results (Tie et al. 2022).Prognostic effects of ctDNA in gastroesophageal cancer patients have recently been published and feasibility as a widely used prognostic markers is under evaluation (van Velzen et al. 2022;Shoda et al. 2017;Iqbal et al. 2019).Moreover, trials investigating further systemic treatment in patients with high ctDNA levels in gastroesophageal cancer are underway and might shed some light on the association of ctDNA with systemic inflammatory parameters such as mGPS.
Furthermore, it was hypothesised that systemic inflammatory biomarkers might pose as predictive parameters for response to immunotherapy.A recently published retrospective analysis of hepatocellular carcinoma patients undergoing checkpoint inhibition suggests that NLR and PLR are independent negative prognostic factors (Muhammed et al. 2021).Similar associations could be drawn in other tumour entities and with various immunotherapeutic agents (Stares et al. 2022;Ye et al. 2023).Yet, so far, only tumour tissue-based markers, such as PD-L1, tumour mutational burden (TMB-H) and mismatch repair deficiency/microsatellite instability (dMMR/MSI-H), are approved biomarkers for administration of immunotherapy (Food and Drug Administration 2017a, b, 2020, 2021).First attempts to correlate these markers with soluble counterparts form liquid biopsies showed promising potential (Mair et al. 2020;Shin et al. 2023).In addition, a post hoc examination of clinical data from the CheckMate-648 trial suggested a potential association between baseline NLR and different response rates in esophageal squamous cell carcinoma patients receiving immunotherapy, chemotherapy or immuno-chemotherapy (Chau et al. 2022).However, data on gastroesophageal adenocarcinoma are still scarce as immunotherapy has only been approved within the last few years.This is underlined by the relatively small subgroup of patients who were tested for PD-L1 or dMMR/MSI-H expression in our cohort.
Furthermore, patients without PD-L1, TMB-H or dMMR/ MSI-H tumours in some cases also show exceedingly promising responses leading to the hypothesis that other biomarkers might provide more sufficient predictive value (Zhang et al. 2021).Thus, further research is warranted to investigate the predictive potential of systemic inflammatory biomarkers in respect to immunotherapeutic strategies.
Due to the retrospective design of our analyses, further exploration of associations with immuno-histochemical staining, such as PD-L1 and dMMR/MSI-H, as well as underlying infectious or autoimmune diseases, therapeutic strategies and longitudinal evaluation of systemic inflammatory parameters could not be performed and therefore pose limitations to our work.In addition, the large timespan of patient recruitment poses a potential bias and is included in our analyses.However, year of first diagnosis was only significantly associated with the OS in the overall but not in the stage-based subgroups as throughout the last decades, more severe cases have been transferred and treated at our department.Despite these limitations of the analysis, our results pose an important cornerstone for further evaluation of mGPS in gastroesophageal adenocarcinoma patients.
In conclusion, higher systemic inflammatory biomarkers were statistically significantly associated with more severe tumour stages and weight loss at first diagnosis.Concerning subgroups based on tumour stages, mGPS showed robust prognostic value in patients with locally advanced disease, which is in line with results from Asian cohorts.Further analyses are warranted to explore the cause and potential clinical applications of elevated mGPS scores.

Fig. 1
Fig. 1 Kaplan-Meier overall survival (OS) estimates for modified Glasgow Prognostic Score (mGPS) in the overall cohort (A), localised setting (B), locally advanced setting (C) and advanced or metastatic setting (D)

Table 1
Systemic inflammatory ratios in a cohort of 769 patients with gastroesophageal adenocarcinoma Abbreviations: NLR neutrophil-lymphocyte ratio, LLR lymphocyte-leucocyte ratio, PLR platelet-lymphocyte ratio, SD standard deviation, HR hazard ratio, CI confidential interval