Absolute Monocyte and Platelet Counts May Provide Additional Prognostic Information in Primary Gastric Diffuse Large B-cell Lymphoma Patients Treated with Rituximab and CHOP

Introduction: Primary gastric diffuse large B cell lymphoma (PG-DLBCL) is the most common histological subtype of primary gastric lymphoma. The standard of care of PG-DLBCL patients is the combination rituximab-based immunochemotherapy (R-CHOP). Recently, different host-related factors have been shown to have significant prognostic significance in non-Hodgkin lymphoma. However, data regarding their prognostic contribution to PG-DLBCL are limited. Aim: To assess the prognostic impact of a panel of simple, cost-effective laboratory variables which are easy to apply in routine laboratory use for R-CHOP-treated PG-DLBCL patients in an attempt to identify those among them that are high-risk category. Materials and methods: We retrospectively assessed the possible prognostic impact of different laboratory markers in 42 R-CHOP treated PG-DLBCL patients treated between 2004 and 2014 and followed at a single institution. Results: The estimated 5-year overall (OS) and progression-free survival (PFS) of the whole group were 80.9% and 78%, respectively. The absolute monocyte and platelet counts in univariate analysis predicted PFS and OS when analyzed as continuous and dichotomized variables. On multivariate analysis performed with factors included in the stage-modified International Prognostic Index (m-IPI), the absolute monocyte and platelet counts remained independent predictors of PFS and OS. Therefore, the absolute monocyte and platelet counts were combined to generate a prognostic index that identified patients with an especially poor overall survival. Conclusions: This prognostic index was independent of the m-IPI and could provide additional prognostic information for better stratification of these patients.


INTRODUCTION
Primary gastrointestinal involvement is one of the most common extranodal presentations in non-Hodgkin lymphoma (NHL), accounting for 4-20% of all NHL cases. 1,2 The most commonly affected site of the gastrointestinal tract is the stomach, followed by the small and the large intestines. 3 Primary gastric lymphoma is accounting for 5% of primary gastric malignancies and the most common histological subtype is diffuse large Bcell lymphoma (PG-DLBCL). 4 PG-DLBCL has been treated with various modalities, including surgery, chemotherapy and radiotherapy alone or in combination with other modalities. [5][6][7] Presently, the combination chemotherapy with cyclophosphamide, doxorubicin, vincristine, prednisone, and rituximab (R-CHOP) is the standard treatment for PG-DLBCL patients. 8,9 However, with improved outcomes attributed to the rituximab addition, the identification of a high-risk patients' subset with an anticipated 5-year survival of less than 50% remains a challenge. Gene-expression profiling, immunohistochemistry-based detection of prognostic biomarkers and early interim analysis with positron emission tomography following the initiation of immunochemotherapy have all been explored as predictors that may identify high-risk patients. [10][11][12][13][14][15][16] Although promising, many of these methods are costly, difficult to obtain, and not easily interpreted. Therefore, it's worth it to identify clinically relevant prognostic factors that are inexpensive, widely available, and easily interpreted by clinicians involved in the PG-DLBCL patients' care.

AIM
The purpose of the study was to assess the possible prognostic impact of a panel of simple, cost-effective and applicable for routine clinical use laboratory variables in R-CHOP-treated PG-DLBCL patients in an attempt to identify the high-risk category.

PATIENTS AND METHODS
The electronic database of the Specialized Hospital for Active Treatment of Hematological Diseases -Sofia, Bulgaria was searched for all patients diagnosed with PG-DLBCL between 2004 and 2014 and treated with a combination immunochemotherapy. Forty-two cases were identified. All PG-DLBCL patients underwent esophagogastroduodenoscopy and biopsy samples were obtained. Diagnosis was defined according to the criteria outlined in the WHO classification. 21 The immunophenotype was determined by automated immunohistochemistry (BondMax, Leica Biosystems) and a panel of monoclonal antibodies against CD20, CD5, CD10, bcl-6, MUM1/IRF4, bcl-2, and Ki-67 (Leica Biosystems) was used in all cases.
Whole-body computed tomography (CT) scanning and trephine biopsy were included as lymphoma staging procedures. The stage at presentation was assessed according to the Lugano staging system, and the localized and advanced diseases were defined as stage I-II 1 and stages II 2 -IV, respectively. 22 The prognostic stratification was performed according to stage-modified International Prognostic Index (m-IPI) that included Lugano stage ≥II 2 , age >60 years, elevated serum lactate dehydrogenase (LDH), performance status (PS) ≥2 and ≥1 extranodal site involvement (excluding stomach).
All patients were subjected to 6-8 curative cycles of R-CHOP regimen (day 1: rituximab 375 mg/m 2 , cyclosphosphamide 750 mg/m 2 , doxorubicin 50 mg/m 2 , vincristine 1.4 mg/m 2 and days 1-5: prednisone 40 mg/m 2 ), and each cycle was repeated every 3 weeks. After completion of therapy, evaluation of the treatment response was performed by gastroscopy with biopsy and CT of the thorax and abdomen. Response was defined according to the criteria of the International Harmonization Project. 23 Laboratory levels of SA, Hb, LDH, B2M, ANC, AMC, ALC and PC were recorded at time of diagnosis.

Statistical analysis
A receiver operating characteristic (ROC) curve analysis was used to illustrate in our data set the best cut-off values of SA, B2M, Hb, ANC, AMC, ALC and PC to predict progression-free survival (PFS) and overall survival (OS) using the Kaplan-Meier method. 24 PFS was calculated from the date of diagnosis to the date of disease progression, death or last follow-up. OS was calculated from the date of diagnosis to the date of death or last follow-up. The log-rank test was used to compare subgroups in the univariate analysis of survival. Multivariate analysis of factors related to survival was performed using the Cox's proportional hazards regression method considering variables that were statistically significant according to the univariate analysis. 25 The effect size was reported as a hazard ratio (HR) with the associated 95% confidence interval (95% CI). Comparisons were performed using Fisher's exact test for binary variables and Kruskal-Wallis nonparametric test for continuous variables. P values <0.05 were considered statistically significant. All data were analyzed using SPSS, version 15 (SPSS Inc., Chicago, IL, USA).

Patients' characteristics
The characteristics of 42 PG-DLBCL patients included in the present study are shown in Table 1. The patients' median age was 56.7 years (range, 19.1-76.4 years). Overall response was observed in 78.6 % patients -complete and partial responses were achieved in 76.2% and 2.4%, respectively. Six patients received salvage therapy whereas two other patients with progressive disease underwent autologous stem cell transplantation. One patient refused further therapy. At the time of analysis, 10 patients progressed and 8 of them died. Median follow-up was 36 months for the entire cohort (range, 2.4-135 months) and 49 months for censored patients (range, 13-132 months). The 5-year PFS and OS for the entire cohort were estimated as 78.0% and 80.9%, respectively.

Univariate analysis of the main prognostic variables
Univariate analysis (Cox regression) of the main prognostic factors for OS and PFS was performed. The HR and the confidence interval (95% CI), and statistical significance were determined for each continuous variable ( Table 2). As continuous variables, SA, B2M, ANC, AMC, and PC had a significant impact on OS and PFS (Table 2).
Using data from the entire cohort, we select the most discriminative cut-off values for the SA, B2M, ANC, ALC, AMC, Hb, and PC to predict survival outcome by ROC curve analysis (Tables 3, 4).
AMC values above the cut-off value of 0.518×10 9 /L were observed in 42.9%. No association was established between AMC and other characteristics ( Table 5). After dichotomization by the most discriminative cut-off value, AMC>0.518×10 9 /L predicted for shorter OS and PFS (Figs 1A, 1B). The 5-year OS was 95.7% in patients with AMC≤0.518×10 9 /L compared to 61.1% in those with higher AMC (p=0.004), whereas the 5-year PFS was 91.7% in patients with AMC≤0.518×10 9 /L compared to 55.6% in those with higher counts (p=0.004).

Prognostic significance of platelet counts
The median PC at diagnosis was 339.5×10 9 /L (range, 207 -696×10 9 /L) and 11 patients had a count above the upper normal level (440×10 9 /L). No association between median PC and m-IPI score risk groups was established. As a continuous variable, the PC was significantly associated with OS and PFS ( Table 2). The most discriminative PC cut-off value on the ROC curve was >335×10 9 /L predicting for OS ( Table 3) and PFS ( Table 4), respectively. PC>335×10 9 /L were observed  Table 6).

Multivariate analysis of the main prognostic variables
A backward selection, using a significance level of 0.05 for dropping variables, was employed to build a Cox proporti- onal hazards model (Table 8), starting from the model with all the prognostic variables identified by univariate analysis of OS (number of extranodal sites >1, AMC >0.518×10 9 /L, SA >33.9 g/L, elevated LDH, platelet count >335×10 9 /L and hemoglobin level >105 g/L). The m-IPI score was not included as a specific variable because the significant fac-  The AMC/PC prognostic index identifies poor-risk patients and provides additional prognostic information when superimposed on the m-IPI. We tried to build a simple prognostic index (PI) which does not contribute to the m-IPI score. We combined the dichotomized AMC and PC to generate the AMC/PC PI and stratified patients into three risk groups: low-risk (AMC≤0.518×10 9 /L and PC≤335×10 9 /L), intermediate-risk (AMC≤0.518×10 9 /L or PC≤335×10 9 /L) and highrisk (AMC>0.518×10 9 /L and PC>335×10 9 /L) populations. High risk PG-DLBCL patients identified by AMC and PC had significantly higher rate of advanced (IV) stage of the lymphoma, additional extranodal lymphoma involvement, elevated LDH levels, poor PS ≥ 2, m-IPI (≥2) scores and higher ANC levels (p=0.005; p=0.004; p=0.002, p=0.02, p<0.001, and p=0.004, respectively) ( Table 9).
OS and PFS were analyzed using the AMC/PC PI and the m-IPI. The median OS was not reached for the low and intermediate-risk groups identified by AMC/PC PI with an estimated 5-year OS of 100% and 90.9%, respectively. Similarly, using the m-IPI, the estimated 5-year OS among low-, low/intermediate-and high/intermediate-risk patients was 93.3%, 100%, and 66.7%, respectively. In contrast, the median OS was only 0.4 years (95% CI 0.15-0.89 years), with an estimated 5-year OS of 25% among high-risk patients identified using the AMC/PC PI (Fig. 3a), compared to 1.09 years (95% CI 0.27-1.92 years) median OS and 38.1% 5-year OS in the high-risk group defined by the m-IPI (Fig. 3b).
The AMC/PC PI identified a group of poor-risk patients with a median PFS of 0.33 years (95% CI 0.08-0.59 years) and an estimated 5-year PFS of 25%. Similarly, the median PFS was 0.5 years (95% CI 0.41-0.69 years) and the estimated 5-year PFS was 42.9% among high-risk patients identified using the m-IPI. Clearly, the AMC/PC PI was able to risk-stratify patients in a manner comparable to the m-IPI.
Furthermore, the AMC/PC PI retained statistical significance at multivariate analysis on OS and PFS after controlling for the m-IPI (Table 10).
Therefore, we sought to determine whether it may provide additional prognostic information when combined with the m-IPI. To test this possibility, low-, intermediate-, and high-risk patients were segregated by the m-IPI. Due to the lack of death events, no statistics could be calculated in the low-risk group. The intermediate-risk (high-intermediate and low-intermediate were combined) patients (n=20) identified by the m-IPI were subsequently risk stratified using the AMC/ PC prognostic index (Fig. 4).
In this group, the 5-year OS was 100% among low-and intermediate-risk patients identified by the AMC/PLC PI (Fig. 4). In contrast, 20% of patients identified by the m-IPI as 'intermediate-risk' , upon further risk stratification by the AMC/PC PI, had dismal outcomes, with a median OS of only 4.8 months (95% CI 0-34 months) and 2-year OS of 25% (p<0.001). The 5-year PFS was 100% and 90% among low-and intermediate-risk patients by the AMC/PLC PI (Fig. 5). However, 25% patients upon further risk stratification by the AMC/PC PI, had poor outcomes, with a median PFS of 3.9 months (95% CI 0-29 months) and 2-year PFS of 25% (p=0.003).
Due to the low number of patients, no prognostic sig-

DISCUSSION
The results of the present study indicated that the PC and AMC at diagnosis may be used as biomarkers for predicting survival in PG-DLBCL patients treated with R-CHOP. The elevated PC and AMC correlated with a worse prognosis in the PG-DLBCL and identified high-risk patients. To the best of our knowledge, no other studies assessing the prognostic impact of PC and AMC in PG-DLBCL patients treated with R-CHOP have been published to date. The gold standard for risk stratification and prognosis in patients with PG-DLBCL continues to be m-IPI. However, other prognostic markers have been investigated to  11,26 In contrast, the measurement of AMC and PC is a simple test that could be easily employed in regular clinical practice.
In the present study, the elevated PC was significantly associated with advanced tumor clinical stage (stage IV), additional extranodal lymphoma involvement, lower SA levels, higher ANC and unfavourable m-IPI score. An elevated PC represented an independent and powerful predictor of poor outcome. The prognostic value of PC was retained at multivariate analysis after controlling for the main other adverse prognostic indicators. Our results are in line with data provided by previously reported studies that confirm the adverse prognostic impact of elevated PC in cancer patients. [27][28][29][30][51][52][53][54][55] However, according to other DLBCL studies, thrombocytopenia rather than thrombocytosis is a poor prognostic factor. 20 This could be explained by the different patients' populations and different rates of bone marrow infiltration.
The prognostic impact of PC might be associated with a more aggressive biology related to earlier hematogene-ous dissemination in cases with elevated counts. Platelets activated by tumor cells play major roles in aiding and abetting tumor progression. First, platelets can help tumor cells survive immune surveillance in the blood circulation. Activated platelets may act as protective "cloaks" for circulating tumor cells, shielding them from immune destruction by natural killer cells. 31,32 This process is mediated by platelet-derived growth factor and transforming growth factor β. 33,34 Hematogenous dissemination of tumor cells can also be facilitated by "platelet mimicry, " in which tumor cells acquire a phenotype that closely resembles that of platelets' . In addition, platelets and vascular endothelium may facilitate tumor cell extravasation and seeding through adhesion molecules P-and L-selectins, a hypothesis supported by the observation that tumor metastases are reduced in mice lacking them. 35,36 In the present study, the elevated AMC represented an independent, powerful predictor of unfavourable outcome that is highly consistent with previously published studies in solid tumors and lymphomas. 10,[37][38][39][40][41][51][52][53][54][55] The exact pathophysiology for the association between high monocyte counts and poor prognosis is not well understood. However, being an important player in immunity, monocytes might play a role in the chronic inflammation which has already been associated with the induction of stomach can-cer and the link with cancer progression has been known for over a century. 42,43 Besides, chronic inflammation is associated with suppression of anti-tumor immune responses and a tumor microenvironment that favours tumor progression. So firstly, AMC may relate to the insidious progression of cancer disease involving activation of innate immune system through mobilization of monocytes to tissue macrophages that develops an inflammatory state associated with increased risk of cancer and mortality. 44,45 Monocytes play a key role in innate immunity, constitute nearly 5% of the circulating white blood cell pool, and exhibit a short half-life in the circulation of a few hours. 46 Secondly, the monocytes, known as the key component of inflammation system, might directly stimulate cancer cell growth by producing various proinflammatory cytokines, such as interleukin-1, interleukin-6, and tumor necrosis factor. Tumor-associated macrophages (TAMs) appear to play a crucial role in the tumor microenvironment and can educate and control invading leukocytes to promote angiogenesis, viability, motility, and invasion. 44,45 From a practical point of view, the proposed AMC/PC PI derived from a CBC at diagnosis is widely available and may be easily incorporated into clinical practice. Furthermore, the AMC/PC PI provided prognostic information independently of that included in the m-IPI and was able to provide additional prognostic information when used in conjunction with the m-IPI. This was particularly true for low-intermediate-and high-intermediate-risk patients who collectively accounted for approximately 50% of the patients in our series. When patients identified as intermediate risk by the m-IPI were further risk stratified by the AMC/PC prognostic index, 20% of these patients were identified as high risk.
In addition, our findings might raise certain issues of reassessment of therapeutic approaches in particular patient categories. With drug development, treatment for PG-DLBCL has shifted from surgery to gastric preservation methods, including immunochemotherapy in order to aid organ conservation. The addition of rituximab to CHOP regimen has improved the therapeutic outcomes, and the superiority of RCHOP has been confirmed by several clinical studies. [47][48][49] However, R-CHOP proved to be of limited value in high risk patients stratified by the AMC/PC prognostic index. The evidence linking both platelets and myeloid-lineage cells with hematogeneous dissemination, and in this study with poor PFS and OS, raises the possibility that strategies which reverse elevated PC or deplete tumorigenic myeloid-lineage cells may represent novel therapeutic approaches in DLBCL. Despite this, to date no specific platelet function targeted therapy has been included in standard cancer treatment. Therapeutic strategies targeting myeloid-lineage cells, including myeloid-derived suppressor cells, are currently being investigated in solid tumors and may warrant further investigation in NHL. 50

CONCLUSIONS
In the present study, we found that pretreatment routine hematological parameters including PC and AMC correlated with prognosis in PG-DLBCL patients treated with R-CHOP. Although this study has certain limitations since it was a retrospective analysis and a single-center study, the AMC/PC PI demonstrates that variables related to the host immune response and the tumor microenvironment are worth consideration when generating prognostic indices in NHL, as these variables may provide additional prognostic information independently from conventional patient-and tumor-specific factors. The low cost, easy accessibility and reproducibility of CBC are other features promoting its use in clinical practice. To confirm these findings, larger, prospective, randomized studies are required in future.