Comparison of the Prognostic Value of Inflammation-Based Scores in Patients with Hepatocellular Carcinoma After Anti-PD-1 Therapy

Introduction

Hepatocellular carcinoma (HCC) is the most common types of liver cancer and the third leading cause of cancer-related deaths worldwide.¹ Due to its uneventful onset and rapid progression, most patients with HCC fail to meet the criteria for radical resection.² In recent years, research on programmed cell death protein-1 (PD-1) inhibitors has continued to emerge, offering new treatment models for HCC patients.³ Given that nivolumab and pembrolizumab have shown certain efficacy and safety, anti-PD-1 therapy is commonly used as a treatment option for systemic therapy.^4,5 However, the efficacy of immunotherapy varies greatly among individuals. Practical and reliable prognostic predictors are needed in anti-PD-1 treatment.

Inflammation is a hallmark of cancer.⁶ Emerging evidence shows that the host inflammatory response is related to cancer progression and patient survival.^7,8 In immunotherapy, inflammation has a predominant role in tumor survival and proliferation, angiogenesis and immunosuppression.⁹ Recently, many inflammation-based prognostic scores composed of systemic inflammatory response factors, including platelet-to-lymphocyte ratio (PLR),¹⁰ neutrophil-to-lymphocyte ratio (NLR),¹¹ lymphocyte-to-C-reactive protein (CRP) ratio (LCR),¹² lymphocyte-to-monocyte ratio (LMR),¹³ systemic immune inflammation index (SII),¹⁴ CRP-to-albumin (ALB) ratio (CAR),¹⁵ prognostic nutritional index (PNI),¹⁶ Glasgow Prognostic Score (GPS),¹⁷ modified Glasgow Prognostic Score (mGPS),¹⁸ and prognostic Index (PI),¹⁹ were proposed to exhibit good predictive ability in cancer prognosis. High pre- and post-treatment NLR and PLR was reported to be associated with worse tumor response and increasing risk of death for anti-PD-1 treatment in HCC.²⁰ Though, the value of inflammation-based scores in predicting survival after anti-PD-1 treatment in HCC patients has not been fully elucidated.

Herein, this study aimed to make a direct comparison of the prognostic value of various inflammation-based scores in HCC patients after anti-PD-1 therapy.

Materials and Methods

This study was conducted according to the ethical guidelines of the 1975 Declaration of Helsinki. The analysis of patient data was reviewed and approved by the Institutional Review Board and Human Ethics Committee at the Sun Yat-sen University Cancer Center (SYSUCC; Guangzhou, China, B2020-190-01).

Patients

Between July 2018 and December 2019, patients diagnosed with HCC who received anti-PD-1 therapy at SYSUCC were screened for eligibility. Patients were included based on the following criteria: (a) diagnosed with HCC through imaging or pathology according to the American Association for the Study of Liver Diseases (AASLD) practice guidelines;²¹ (b) had confirmed records of receiving PD-1 inhibitors; (c) aged from 18 to 75 years; (d) had a performance status (PS) score less than 2 and a Child-Pugh (CP) stage of A or B; (e) had no other malignant tumors; and (f) had complete medical and follow-up data. All laboratory serum test data were collected within 3 days before the initial use of PD-1 inhibitors. Imaging evaluation included enhanced computed tomography (CT) or magnetic resonance imaging (MRI) examination within a week before the administration of anti-PD-1 therapy.

Treatment Procedure

PD-1 inhibitors were administered intravenously with saline. The types and dosages of drugs are summarized in Table S1. PD-1 inhibitors could be combined with locoregional interventional therapies or tyrosine kinase inhibitors (TKIs) including sorafenib, lenvatinib, regorafenib or apatinib during treatment. Drug discontinuation was applied upon disease progression, the development of unacceptable toxicity, patient withdrawal of consent, or changes in the treatment plan.

Inflammation-Based Prognostic Scores

The counts of white blood cells, neutrophils, lymphocytes, monocytes, and platelets and the levels of CRP and ALB were obtained through routine laboratory tests of blood samples. The PLR, NLR, LCR, LMR, SII, CAR, GPS, mGPS, PI and PNI were calculated as described in Table 1.

Table 1 Systemic Inflammation-Based Prognostic Scores

Follow-Up and Definitions

The final follow-up ended on March 31, 2021. Follow-up examinations were conducted using laboratory tests including serum alpha fetoprotein (AFP), liver function, and routine blood tests. Abdominal ultrasonography, enhanced CT or MRI was performed every 1 to 2 months after the receipt of PD-1 inhibitors. Follow-up intervals were routinely 2 to 4 months.

Tumor response was defined as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD) according to the modified Response Evaluation Criteria in Solid Tumors 1.1 (mRECIST).²² Overall survival (OS) was defined as the time interval from anti-PD-1 initiation to cancer-related death.

Statistical Analysis

The results are described using the median and range for nonnormally distributed values.

Groups were compared by using Student’s t-test for continuous data and the χ² test for categorical data. Survival analysis was performed using the Kaplan-Meier method, and differences in the survival curves were analyzed with the Log rank test. For single value indicators, to avoid deviations from different criteria for the cutoff values of the prognostic scores in this cohort, the optimal cutoff point was calculated using X-tile v.3.6.1 software for PLR, NLR, LCR, LMR, SII, CAR, and PNI (Figure S1). For composite indicators, the GPS, mGPS and PI scores were calculated as generally reported.^17–19 Univariate and multivariate Cox regression analyses were performed to determine prognostic factors for OS. Hazard ratios (HRs) and confidence intervals (CI) were also calculated. All variables with a P-value < 0.05 in the univariate analyses were used in the multivariate analyses using Cox proportional hazards models. A two-tailed P-value < 0.05 was considered statistically significant. Time-dependent receiver operating characteristic (ROC) curves at 6, 12, 18 and 24 months and the area under the curve (AUC) were calculated to compare the predictive ability of the ten inflammation-based scores. Harrell’s concordance index (C-index) was determined to evaluate the predictive ability of the ten inflammation-based scores. All data analyses were performed using SPSS 25.0 software (SPSS Inc., Chicago, IL), GraphPad Prism (version 8.0; GraphPad, Inc.) and R version 4.0.2.

Result

Identification and Characteristics of the Study Patients

From July 2018 to December 2019, a total of 442 patients who received anti-PD-1 therapy and met the criteria were included in the study. Among them, 382 (86.4%) were males. Ages ranged from 21 to 75 years with a median of 52 years. Consistent with the characteristics of HCC in China, 372 (84.2%) patients had hepatitis B. A majority of patients were classified as CP stage A (92.8%) and BCLC stage C (72.2%), and most had multiple tumors (67.9%). A total of 232 (52.5%) patients had macrovascular invasion, and 176 (39.8%) patients had extrahepatic metastasis. The range of tumor size is from 1 to 21.5 cm, with a median of 8.9 cm. The clinical characteristics, including the ten inflammation-based scores of the patients, are summarized in Table 2.

Table 2 Baseline Characteristics of the Enrolled Patients

Of note, most patients received another antitumor treatment before anti-PD-1 therapy, including surgery (14.0%), transcatheter arterial chemoembolization (TACE) (20.4%), hepatic infusion chemotherapy (HAIC) (21.0%), ablation (8.1%) and TKIs (11.3%). The duration of anti-PD-1 therapy ranged from 1 to 18.9 months with a median of 5.7 months. The number of treatment cycles ranged from 1 to 25 with a median of 5. The types of PD-1 inhibitors included nivolumab (6.6%), pembrolizumab (7.7%), toripalimab (62.7%), sintilimab (21.3%) and camrelizumab (6.3%) (Table S1). The median duration of follow-up was 13.7 months, while the median OS was 18.3 months. A total of 204 (46.2%) patients died, while 195 (44.1%) patients survived at the end of the follow-up.

Independent Prognostic Factors for OS

Univariate analysis involved prognostic factors related to ethnic characteristics, liver function, tumor burden, the treatment process and the ten inflammation-based scores.

In addition to liver function and tumor stage, all ten inflammation scores were identified as significant prognostic factors for OS. Multivariate analysis revealed that total bilirubin (TBIL) (P = 0.025), Barcelona Clinic Liver Cancer (BCLC) stage (P = 0.003), largest tumor size (P < 0.001), tumor number (P = 0.001) and PNI (P < 0.001) were independent prognostic factors for OS (Table 3).

Table 3 Univariate and Multivariate Time-Dependent Cox Regression Analyses of the Prognostic Factors for Overall Survival

Overall Survival

As shown in Figure 1, all inflammation-based scores were associated with the OS of patients who received anti-PD-1 therapy. High PLR, NLR, LCR, LMR, SII, CAR, GPS, mGPS, PI and PNI scores indicated poor prognosis (all P < 0.01). However, only the PNI score remained significant in multivariate analysis. The PNI score differentiated the HCC patients into two groups with distinct prognoses (6-, 12-, and 24-month OS rates: 90.9%, 71.7%, and 51.5% and 77.6%, 50.7%, and 25.7%, respectively; P < 0.001).

Figure 1 Kaplan-Meier curves of the overall survival of HCC patients after anti-PD-1 therapy. (A) PLR, (B) NLR, (C) LCR, (D) LMR, (E) SII, (F) CAR, (G) GPS, (H) mGPS, (I) PI and (J) PNI.

Comparison of the Performance of the Inflammation-Based Scores

Time-dependent ROC curves at 6, 12, 18 and 24 months of OS were formed to compare the performance of the ten inflammation-based scores (Figure S2), and the PNI score was superior to the others. The AUC of the time-dependent ROC curve showed that the PNI score had a better ability to predict OS. Plots of the time-dependent AUC are shown in Figure 2. The C-indexes were calculated, and the values are provided in Table 4. The PNI scores consistently had higher C-index values than the other scoring systems.

Table 4 Concordance Index for the Comparison of Different Inflammatory-Based Scores

Figure 2 Time-dependent AUC plot for survival prediction of inflammation-based scores.

Relationships Between the PNI Score and Clinical Characteristics and Efficacy

The correlations between clinical characteristics and efficacy and the PNI score are shown in Table 5. A high PNI score was associated with worse aspartate aminotransferase (AST) (P < 0.001), ALB (P < 0.001), TBIL (P < 0.001), and Child-Pugh grade (P < 0.001); increased CRP (P < 0.001) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) (P = 0.011); and larger tumor size (P < 0.001), multiple lesions (P = 0.020), and macrovascular invasion (P = 0.004). Tumor response analysis indicated that the objective response rate (ORR) (34.9% vs 21.1%, P = 0.001) and disease control rate (DCR) (75.5% vs 59.6%, P < 0.001) were higher in the low PNI group.

Table 5 Baseline Characteristics of the Patients Grouped by PNI Score

Discussion

It is widely acknowledged that inflammation-based scores are associated with cancer-specific survival, but the inflammatory biomarkers that best predict prognosis in anti-PD-1 therapy, especially in HCC patients, remains unclear. To the best of our knowledge, this is the first study to comprehensively identify the correlation between inflammation scoring systems and the OS of HCC patients who received anti-PD-1 therapy. Moreover, the PNI score was found to be superior to the other inflammation-based scores in prediction.

HCC is an inflammation-driven carcinoma, for the majority of HCC arises in the context of chronic inflammation and within a fibrotic liver.^23,24 Emerging evidence has shown that the inflammatory response is correlated with the efficacy of anti-PD-1 therapy in advanced cancers.^9,25,26 The unique immunobiology of the liver under conditions of fibrosis and chronic inflammation presents an opportunity for therapeutic targeting with immune checkpoint inhibitors, such as PD-1 inhibitors.²⁴ A study proposed a tumor inflammation signature constructed with 18 genes that accurately predicted the benefit of patients treated with PD-1 inhibitors.²⁷ However, the practicability of the model is insufficient in clinical practice. For HCC, the NLR and PLR were reported to have strong predictive roles in anti-PD-1 therapy,^20,28,29 which was verified in our study. With the combination of albumin and lymphocytes, our study suggested that the presence of the systemic inflammatory response revealed by PNI was the optimal tool in the assessment of survival in HCC patients treated with PD-1 inhibitors compared with the other inflammation-based scores. The PNI score is measured using an inexpensive, easily available, simplified approach in clinical practice to stratify HCC patients with anti-PD-1 therapy into different risk groups.

The PNI score has been demonstrated to be an effective prognostic predictor in many types of digestive system cancers, including HCC.³⁰ The underlying mechanism has not been well clarified. In this study, one possible explanation is that the decrease in lymphocytes indicates a weakening of immunity.³¹ In addition, the depletion of lymphocytes, such as CD4- and CD8-positive T cells, is the key to immunosuppression.³² Lymphocytes within the tumor microenvironment seem to be critical in determining the efficacy of immune surveillance and lethality.³³ On the other hand, albumin reflects liver function, which determines the sustainability of immunotherapy and prognosis in HCC patients. Albumin is an indicator of nutritional status. A study indicated that nutrition and metabolism in patients with advanced HCC were closely related to anti-PD-1 treatment efficacy as well as survival benefit.³⁴ Low serum albumin levels reflect a state of malnutrition, which would weaken cellular and humoral immunity, phagocytic functions, and other defense mechanisms in patients with cancer.³⁵

This study has some limitations. First, this was a retrospective study conducted on a single-center cohort in China. A majority of the included patients had hepatitis B-related HCC. Since hepatitis C virus infection, alcohol abuse and an unbalanced diet are the leading causes of HCC in Japan, the United States, and European countries, the results need to be validated in external institutions with different disease backgrounds. Second, we used measures to control the consistency of each inflammation-based score; however, the best cutoff value of PNI in this study is not widely accepted (for the most common cutoff value is 45). If necessary, it should be revalidated and redefined in future studies with authoritative sources. Third, patients received non-single PD-1 inhibitors during the whole treatment, which inevitably caused bias. Finally, the underlying potential regulatory mechanism of albumin and lymphocytes on immunotherapy has not been elucidated and deserves further investigation.

Conclusions

Our study demonstrated that the PNI score was an independent prognostic indicator in HCC patients receiving anti-PD-1 therapy and performed well compared with the other inflammation-based scores. This is an easy-to-use tool for risk stratification and benefits physicians in the selection of therapeutic strategies for anti-PD-1 treatment in HCC patients.