Is platelet-lymphocyte ratio (PLR) a predictor of thrombosis and together with circulating tumor cells capable to determine recurrence-free survival in patients with gastric cancer?

Cancer associated thrombosis (CAT) is a major cause of morbidity and mortality in cancer patients.1 The risk of venous thromboembolism is 4.1 higher in oncology patients compared to those without cancer;2 idiopathic venous thromboembolism (VTE), which is composed of deep vein thrombosis (DVT) and pulmonary embolism (PE), is associated with 20% of further diagnose of malignant disease.3 Tumor type, stage and extent of the cancer and antineoplastic regimen influence the incidence of CAT,4 but there are no accurate clinical algorithms to identify cancer patients at high risk for VTE.

While the majority of cancer patients remain at low risk for VTE, the identification of patient candidates for surveillance or thromboprophylaxis remains a daunting clinical challenge.5 The development of accurate risk assessment tools to stratify VTE risk in cancer patients had been attempted previously. The Khorana score is calculated by five validated variables: site of cancer; platelet count; hemoglobin level; leukocyte count; and body mass index.6 Two studies sought to improve the predictive value of the Khorana score by incorporating additional variables. The Protech score included treatment with cisplatin, carboplatin and/or gemcitabine,7 whereas the Vienna prediction score added biomarkers of platelet and coagulation activation (P-selectin and D-dimer, respectively).8 Recently, another score (Indicate) was published, which evaluates albumin and LDH levels to predict risk of thrombosis.9

Maybe, variables not included in these tools may influence the risk of VTE. The discovery of additional factors associated with CAT is a pivotal step for the refinement of risk assessment strategies. The level of circulating tumor cells (CTCs) is a prognostic biomarker of progression-free survival and overall survival for many solid tumors.10 We recently correlated CTC counts with prognosis in patients with non-advanced gastric cancer.11 In addition, a preclinical study suggested that CTCs may promote VTE,12 and two clinical studies associated CTCs with an increased risk of VTE in metastatic breast cancer patients.13 14

CTCs aggregate with platelets and coagulation factors to form circulating tumor microemboli (CTM) that are more likely than CTCs to overcome the stressors of physical shear forces and immune surveillance in the bloodstream.15 In addition to facilitating hematogenous metastasis, CTM may activate the coagulation cascade through the interactions of platelets, tissue factor, fibrin, and selectin.16 Consequently, CTM could link CTCs to the pathophysiology of CAT. However, a possible association of CTM with VTE has not been evaluated in patients with advanced neoplasms.

Clinicians currently include platelet counts in Khorana score calculations, but do not incorporate the platelet-lymphocyte ratio (PLR). PLR is a marker for poor prognosis in coronary artery disease.17 18 In a cohort PLR was associated with higher VTE incidence in an ambulatory cancer population.19 The aim of the present study was to evaluate the correlations of CTCs, CTM, and PLR with VTE and assessing these variables relationships to recurrence-free survival (RFS).

We conducted a prospective single-center study at the A.C. Camargo Cancer Center, São Paulo, Brazil. Patients with gastric cancer were recruited at the Department of Abdominal Surgery, from March 2016 to April 2017, and were followed until January 2018. This study was approved by the institutional research ethic committee (Protocol No. 2134/15).

Inclusion criteria were diagnosis of gastric adenocarcinoma; age >18 years; measurable or evaluable disease; and no surgery for <4 weeks prior to sample collection. Patients receiving therapeutic anticoagulation were excluded. Methods of CTC analysis of patients with non-metastatic gastric cancer were published recently.10 After obtaining written consent, blood samples for CTC assays were collected before the start of the first cycle of neoadjuvant chemotherapy (usually FOLFOX or XELOX) for patients with locally advanced tumors, and prior to the first cycle of first-line chemotherapy for patients with metastases. The second CTC evaluation was completed after surgery or before adjuvant treatment for patients with localized disease, and after 6 months of treatment for patients with metastases. The incidence of VTE was the only variable determined by retrospective review of electronic medical files. The VTE examinations were performed by clinicians when patients were symptomatic and the asymptomatic cases where incidentally found. Data regarding age, tumor histology, and metastasis were recorded and analyzed for their associations with VTE. We also evaluated complete blood counts, liver function, and serum tumor markers collected at the clinical analysis laboratory of AC Camargo Cancer Center. PLR was evaluated only at baseline, due to difficulties in obtaining data from medical records. We estimated PLR cut-off point for VTE using receiver operating characteristic curve (ROC curve). We established 288 because it was the point of high specificity, improving positive predictive value (Figure 1). PLR cut-off point (297) for RFS was estimated by using the maximum of the standardized log-rank statistic proposed by Lausen and Schumacher (1992).20

Patients were stratified for VTE risk by using a predictive model for chemotherapy-associated thrombosis (Khorana score), which includes the following variables: site of cancer; platelet count; hemoglobin level; leukocyte count; and body mass index.6 We chose the Khorana score because it is a simple and validated method, indicated for outpatients under chemotherapy. This score was recently included in the ASCO Clinical Practice Guideline Update for VTE prophylaxis in patients with cancer.21

Because a previous study suggested an association of CTCs with an increased risk of VTE in breast cancer patients (12), and due the procoagulant potential of CTM, our rationale was to determine if CTCs/ CTM levels together with PLR could more accurately predict VTE incidence in patients considered at intermediate or high risk according to the Khorana score.

We found a cumulative VTE incidence of 20.4%, which is consistent with the literature. An epidemiologic study of a gastric cancer population found a 2-year cumulative VTE incidence that ranged from 0.5% to 24%, varying according to tumor stages, from I (M0) to IV (M1).23 We found no difference in VTE incidence between CTM-positive or negative groups. As Khorana scores, metastatic disease, and cancer stage could be associated with VTE incidence, we conducted a multivariate analysis, which also showed that VTE incidence persisted unrelated to CTM-positive group. These results suggest that the prediction of VTE will require a complex model that incorporates multiple variables.24

Although there is rationale to suggest a hypothetical relationship of CTM with the incidence of VTE, this potential association was not empirically validated in our clinical setting. Interestingly, patients with high- and intermediate-risk Khorana scores also showed no statistical difference of VTE incidence. Probably, the fact that we analyzed these factors in patients with localized and metastatic disease had interfered with the results.

Our finding of a 50% probability of VTE when PLR is >288 supports the role of platelets in activating the coagulation cascade. This is a strong finding as it correlated with VTE even in a mixture patient population. In addition, higher baseline PLR was also associated with poor PFS, corroborating previous findings that platelets enable CTCs to evade immune responses and facilitate epithelial-mesenchymal transition.16 Thus, our findings suggest that CTCs at the second assessment (CTC2), PLR, and VTE have roles in metastasis, leading to treatment failure and poor RFS.

An interesting finding was that patients with CTCs at baseline had better PFS. We suggest that the early presence of these cells in the bloodstream stimulated effective anti-tumor immune responses. More interesting is the finding that patients with higher CTC levels at CTC2 had poor RFS. We suggest that CTC bloodstream invasion at varying timepoints may have differential effects on RFS.

These results underscore the complexity of CTC, CTM, and platelet interactions and the difficulty to predict which cancer patient will develop VTE.23 Meanwhile, in our population, disease progression was strongly correlated with VTE, which reflects the need for effective thromboprophylaxis. This study highlights that a solution to this conundrum could be the development of safer anticoagulants. Recently, two trials not specific for gastric cancer compared apixaban and rivaroxaban with placebo for CAT thromboprophylaxis.25 26 Lower doses of both rivaroxaban and apixaban seemed safe. Moreover, in the rivaroxaban trial, patients who received pharmacological prophylaxis had a lower, although not statistically different, mortality rate. Unfortunately, direct oral anticoagulants, especially rivaroxaban and edoxaban, may increase the bleeding risk in upper gastrointestinal cancers.27

The burden of CAT imposes not only mortality, but also morbidity, anti-coagulation costs, and anti-neoplastic treatment interruption.28 A better predictive tool is mandatory. Most factors included in prediction models are static, whereas the risk of thrombosis is dynamic during the patient's life span and may be determined simply by chance. We found that the PLR associated not only with poor prognosis, but also with a higher incidence of VTE. PLR could potentially further improve the prediction of VTE. Therefore, PLR could be a new biomarker for VTE risk stratification in the oncology setting.

The drawbacks of our study are first a limited number of patients with localized and metastatic gastric cancer. In our cohort, 86 patients had blood collection for CTC1 analysis and 45 for CTC2. Probably, there was a survivor bias and consequently, selection of patients with better prognosis in the metastatic group. It is possible that most of patients did not make the second blood collection (CTC2) for poor prognosis or death. CTC1 and CTC2 had different timepoints depending on whether the disease was localized or metastatic and this could influence RFS as also the presence of CTM and VTE. Although PLR association with VTE was not our primary outcome, this finding is congruent with previous studies. Besides, our cut-off (288) value was quite similar to Ferroni cut-off (260).19 In the metastatic group when PLR >288, 5 of 8 patients developed VTE, although statistically significant (p=0.048), this group contains limited number of patients. Further studies are necessary to confirm this result.

To the best of our knowledge, our study is one of the very first to find that PLR, CTC2, and VTE are independent prognostic factors for RFS in gastric cancer. Our findings reinforce the difficulty to foresee which cancer patients will develop VTE. Neither CTC, nor CTM improved this prediction, but PLR could constitute new prognostic biomarker with the advantage of being easy, feasible and of low cost. A study of a larger cohort could better evaluate these factors. Until there, the use of safer anticoagulants in patients at low risk of hemorrhage could be continued to address this dilemma.