Prognostic Significance of Elevated Platelet Count (>200 x 10^9 per L) in BCLC Stages B and C of Hepatocellular Carcinoma: A Retrospective Multicenter Analysis

Introduction

Hepatocellular carcinoma (HCC) is a common and deadly cancer, frequently arising in patients with chronic liver disease or cirrhosis.^1,2 Treatment is often complicated by portal hypertension-related comorbidities, contributing to end-stage liver disease symptoms such as ascites, kidney failure, variceal bleeding, and thrombocytopenia.^3,4 Platelet count, a critical marker of significant portal hypertension, is effective in predicting postoperative outcomes in BCLC Stage A patients and guiding clinical decisions.^5,6 Despite increasing data on the relevance of platelet count across other BCLC stages its role has to be further refined.

Early studies have shown that in BCLC Stages B and C, the prognostic role of platelet counts exhibits an opposite trend compared to BCLC Stage A.^7,8 Recent evidence complementing these findings by Scheiner et al, showed in a large patient cohort that increased mean platelet volume and low platelet count correlate with improved survival in HCC patients.⁹ Similarly, Huo et al reported that elevated platelet counts across all BCLC stages (0-D) are associated with advanced tumor characteristics and predict worse overall survival in HCC patients¹⁰ and Chen et al demonstrated a U-shaped curve associating platelet count with survival in BCLC Stage B patients.¹¹ These studies highlight the complex role of platelet counts in HCC progression across different BCLC stages. However, the precise role of platelet counts remains unclear and a clinically relevant cutoff for BCLC B and C patients still needs to be established.

Non-invasive methods, such as CT-derived liver and spleen volumes, are increasingly being used for the evaluation of portal hypertension in HCC.¹² Whereas liver stiffness measurements may not reliably assess clinically significant portal hypertension in HCC patients.¹³ These imaging approaches underscore the utility of certain Baveno criteria elements, particularly spleen diameter, in the non-invasive assessment of portal hypertension in HCC patients.

Delineating the relationship between platelet count, portal hypertension, and additional underlying factors is important for prognostic assessment, improvement of treatment strategies and new therapeutic approaches in HCC. Our study focuses on evaluating platelet count comprehensively across BCLC stage to validate previous findings and to find better clinically meaningful cutoffs for platelet count at diagnosis of HCC.

Methods

Study Design

Epidemiological data, survival, and recurrence rates were analyzed in a cohort of patients diagnosed with HCC between 2006 and 2022. Consecutive Patients were included with following criteria:¹ age ≥18 years, and² a radiological or histological diagnosis of HCC. The study included patients from the University Medical Center Munich, University Clinic Mannheim, and University Clinic Vienna. Data collection and retrospective analysis were conducted in compliance with the Declaration of Helsinki and approved by the local ethics committee (LMU Ethics Committee: 18–604). Patient consent was not required due to the study’s retrospective design. To ensure data confidentiality, all patient information was anonymized throughout the study.

Background and Data Collection

We conducted a retrospective analysis of 1112 patients with HCC from three centers: Munich (n = 754), Vienna (n = 250), and Mannheim (n = 108). In total 632 deceased during the study period while 480 were censored. Diagnosis was made either histologically (postoperatively or by puncture) or by imaging and typical contrast enhancement on the basis of progressed liver disease. Most current lab values available at time or shortly before initial treatment (<3 months) were used for the analysis. Macrovascular invasion was diagnosed based on contrast-enhanced imaging (CT and/or MRI) following the standard diagnostic criteria for HCC. Baseline characteristics assessed included age, sex, height, weight, performance status (according to the Eastern Cooperative Oncology Group), presence of cirrhosis, Child-Pugh score, etiology of liver disease, presence of ascites, portal hypertension, spleen diameter, BCLC Stage and laboratory values. Data were extracted, recorded, and entered into a central database by trained personnel. Information on patient survival and disease recurrence was obtained from clinical reports, death certificates issued by local public health departments, and registration offices. Data processing and storage adhered to the Bavarian Law of Cancer Registries.

Exploratory Analysis of Platelet Range for HCC Patients

To find optimal cut-off values we tested the Log rank test with different cut-offs and chose the cut-offs with the lowest p-value as previously described.^14,15 Patients for which lab values were not available according to the requirement (<3 months prior to first treatment) were not considered for subsequent analysis. To determine the optimal platelet range for patients with hepatocellular carcinoma, we considered the lowest (11 × 10^9/L) and highest (1073 × 10^9/L) platelet counts in our cohort. We iteratively tested various platelet ranges from X to Y x 10^9/L including combinations such as 10–12, 12–14, 100–150, 400–700, until all possible range sizes were covered. Patients were categorized as either inside or outside each range. An exhaustive analysis was conducted by plotting survival curves for all platelet ranges and assessing the separation of these curves. For each range, we fitted a survival model, visualized survival curves, and used the Log rank test to evaluate differences in survival probabilities. The p-value from the Log rank test was computed for each range. The range associated with the best survival outcomes allowed the classification of patients into three groups: “normal range”, “low platelet counts”, and “high platelet counts”, which differed from conventional platelet ranges.

Statistical Analysis

Continuous data were summarized using means, medians, minimum and maximum values, and standard deviations, while categorical data were presented as absolute frequencies and relative percentages. Comparisons of patient characteristics were made using analysis of variance (ANOVA) or the Kruskal–Wallis test for continuous data and Chi-square tests for categorical variables. Survival analysis was conducted from the date of diagnosis until death or the last patient contact. Follow-up and survival times were censored as of December 31, 2022. Univariable analysis involved fitting a series of Cox proportional hazards models to assess the relationship between individual variables and survival. Significant variables were identified, and a multivariable Cox regression model was developed to determine independent predictors of survival. Rows with missing values were excluded to ensure robust analysis. Stepwise forward selection was used to refine the model, starting from a null model and progressively adding variables. The final model summarized significant predictors of survival, with a significance level set at 0.05. All analyses were performed using RStudio 2023.09.1+494 “Desert Sunflower” release.

Results

In this retrospective cohort study, we analyzed a total of 1112 patients diagnosed with HCC from all BCLC stages across three European centers. Baseline characteristics are provided in Table 1.

Table 1 Baseline Characteristics of the HCC Patients Across BCLC Stages

To identify clinically relevant platelet cutoffs, platelet count prognostic potential was examined across BCLC stages. Based on previous observations, we hypothesized that two cutoffs better discriminate survival of HCC patients. To explore the prognostic potential of cutoff pairs, we iteratively tested cutoff pairs that stratified risk of patients in low, medium, and high platelet count groups. A cutoff pair dividing the patients into low (<84/nl), medium (84–200) and high platelet counts (>200/nl) showed the strongest predictive potential (Figure 1A). This cutoff pair showed a significantly better model fit compared to any single cutoff (spp. Table 1). Patients with platelet counts within 84 to 200 x 10^9/L at diagnosis were significantly associated with better survival, whereas platelet counts outside this range were linked to poorer survival (Figure 1B).

Figure 1 (A) Heatmap depicting the statistical significance (z-axis) of various cutoff pairs tested with the Logrank survival test (z-axis), hereby the best survival discrimination was found for an upper cutoff of 200000 platelets/µL and lower cutoff at 84000 platelets/µL (B) Kaplan-Meier survival analysis with newly defined platelet count cutoffs, comparing survival probabilities of patients within the new platelet range (84–200 × 10^9/L) vs outside ranges (Logrank, p<0.0001). Abbreviation: BCLC, Barcelona Clinic Liver Cancer Staging system.

As shown in Table 2, low platelet counts were significantly associated with a higher prevalence of cirrhosis (96.1% vs 84.8% and 56.3%, p < 0.001), more advanced liver disease (reflected by higher Child-Pugh scores), as well as higher incidences of metastasis, macroscopic portal vein infiltration, ascites (50.2%), and portal hypertension (90.3%) (p < 0.001). On the other hand, platelet numbers above 200/nl were significantly associated with BCLC C stages.

Table 2 Baseline Characteristics of Patients According to the Newly Defined Platelet Ranges

According to the new platelet cutoffs we performed a multivariate Cox Regression analysis across BCLC stages and for each BCLC stage separately (Table 3 and Figure 2). Spleen diameter was included in the model to adjust for portal hypertension and to adjust for potential confounders. We further included macroscopic portal hypertension or metastasis. The multivariate analysis showed that in BCLC A patients lower platelet counts (<84/nl) were associated to poorer survival whereas in BCLC B and C patients primarily elevated platelet counts (>200/µL) were associated to poorer survival. Due to the incomplete assessment of varice status (available in 868 of the 1112 patients) it was not included in the previous model. Upon inclusion of varice status the overall model the performance does not improve (Concordance with varices 0.72 versus 0.72 without varices) and the multivariate Cox-Regression analysis results remain nearly unchanged spp Table 2. We also assessed the effect of antiplatelet medication (data available for 593 patients, with 112 receiving treatment); the hazard ratio was 1.14 (p = 0.391), and the concordance index remained 0.72 (spp Table 3). Bootstrapping, performed with 1000 resamples, validated the robustness of the model’s coefficient estimates and the predictive accuracy of platelet counts as a prognostic marker (spp Table 4).

Table 3 Multivariate Cox Regression Analysis by BCLC Stages

Figure 2 (A) Kaplan-Meier survival analysis with the newly defined platelet count cutoffs for BCLC stage A patients (B) Kaplan-Meier survival analysis with the newly defined platelet count cutoffs for BCLC stage B patients (C) Kaplan-Meier survival analysis with the newly defined platelet count cutoffs for BCLC stage C patients. Abbreviation: BCLC, Barcelona Clinic Liver Cancer Staging system.

Discussion

Low platelets count in HCC has been extensively researched as surrogate of portal hypertension, particularly in patients with early-stage disease (BCLC A and B). Thrombocytopenia is considered in this setting a negative predictor of the outcome of surgical or interventional procedures. In contrast, recent research highlighted an association between lower platelet count and favorable outcome in patients with advanced-stage disease (BCLC C and D) (Scheiner et al, 2018). Chen et al recently underscored the association with poor prognosis of elevated platelet counts for BCLC B patients as well.¹⁶ However, only few studies have linked elevated platelet counts to shorter survival.^7,8,16 The findings of Scheiner et al and Chen et al prompted us to examine the precise role of platelet counts in HCC across different BCLC stages, adjusting for the effects of portal hypertension, a factor not considered in previous studies to validate these observations and to find clinically relevant cutoff values.

To further investigate this apparently conflicting information on the prognostic role of platelet counts and given that conventional platelet cutoffs did not adequately discriminate the different populations, we tested various cutoff values for discriminatory potential and found better discriminatory potential upon using two cutoff values (spp. Table 1). We identified that platelet counts above a cutoff of 200/nl indicated poorer survival in the patient population. The multivariate analyses confirmed that elevated platelet counts, according to the new cutoff of 200/µL, were negatively associated with survival in HCC patients. This finding aligns with the results of Pinter et al and Chen et al,^9,16 showing that elevated platelet counts are associated with poorer survival in BCLC B and C stages (Table 3). The new lower cutoff (below 84/µL) closely approximates the commonly used 100/µL cutoff for assessing surgical risk in BCLC A patients hereby validating the study cohort and methodology (Table 3).^5,17,18 A recent meta-analysis further supports our data, demonstrating that while low platelets (<100/nl) predict poorer prognosis in curative-intent HCC patients, elevated platelet counts are associated with worse outcomes in palliative treatment settings (BCLC B and C).¹⁹

Other prognostic markers, in addition to or in combination with platelet counts, can further support prognostic estimation in HCC. Inflammatory markers like the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) are associated with poorer outcomes, especially in HCC with BCLC B stage.^20–24 The aspartate aminotransferase to platelet ratio index (APRI) helps assess liver fibrosis, while C-reactive protein (CRP) reflects systemic inflammation and correlates with advanced disease.^25–27 Lymphocyte-to-monocyte ratio (LMR) and red cell distribution width (RDW) have additionally shown a linkage between hematologic changes to prognosis.^20,28 Additionally, Shi et al demonstrated that integrating platelet counts with CD8+ T cell levels improves prognostic accuracy in HCC patients with portal vein tumor thrombosis (PVTT).²⁹ Combining these markers with platelet counts could create more comprehensive models for risk assessment and treatment planning in HCC.

Preclinical evidence suggests a complex, potentially detrimental role of platelets in hepatocarcinogenesis. For example, platelet-derived GPIbα, which plays a crucial role in platelet adhesion to blood vessels, has been shown to be critical for HCC development in a model of non-alcoholic steatohepatitis.³⁰ Additionally, platelet releasates have been found to enhance the proliferative response of HCC cells.³¹ Other in vitro findings reinforce the importance of platelets in HCC pathogenesis,³² indicating that platelets themselves may play a harmful role. Padickakudy et al reported that platelet-derived serotonin promotes tumor angiogenesis with elevated serotonin levels predisposing patients to early HCC recurrence, collectively linking platelet-driven hypercoagulability to venous thromboembolic events and poorer prognosis.³³ Further studies are needed to delineate the impact of platelets on HCC progression to advance biomarker discovery and therapeutic targeting.

Despite its multicentric nature, our study has several limitations. Platelet counts were not adjusted for inflammatory markers such as C-reactive protein (CRP) or leukocyte numbers, which could influence the results. Additionally, the absence of a validation cohort means that these findings require further confirmation through subsequent studies. Therapy-related variables (eg, sorafenib, SIRT) could further refine prognosis and survival outcomes. Data collection protocols across centers may introduce variability and could impact reliability and standardization. Further the study design is retrospective and missing information (ie variceal status) or inconsistencies could affect the reliability of the conclusions.

The impact of platelet count was adjusted for surrogate markers of portal hypertension and comprehensively analyzed across different BCLC stages. This study revealed a stage-dependent association of elevated platelet counts with poor HCC prognosis and suggests a new platelet cutoff (200/µL) for assessing prognosis in BCLC B and C patients, which could enhance risk stratification and guide therapeutic decision-making.