Impact of neutrophil-to-lymphocyte ratio on survival outcomes among cirrhotic and non-cirrhotic patients with advanced hepatocellular carcinoma under atezolizumab–bevacizumab combination therapy

Introduction

Liver cancer is a major clinical entity and has become increasingly frequent during recent decades, being the 6^th most common type of cancer and the 4^th leading cause of cancer-related mortality globally [1]. The most common type of liver cancer is hepatocellular carcinoma (HCC) [2]. The majority of HCC cases are frequently diagnosed in the intermediate or the advanced stage, where potentially curative interventions are not feasible [3].

In 2007, sorafenib was the first agent to receive approval for the systemic treatment of advanced HCC [4]. Since then, more tyrosine kinase inhibitors (TKIs) have gradually been developed and used for this purpose [5-7]. Recently, the introduction of atezolizumab–bevacizumab in unresectable HCC shifted the balance in the HCC treatment landscape towards new immunological pathways. The IMBRAVE150 study showed a survival benefit in patients with unresectable HCC compared to sorafenib in first-line systemic treatment, yielding objective response rates up to 20%, accompanied by a great improvement in patients’ quality of life [8].

Unfortunately, there are only limited data concerning factors that determine the response to immunotherapy in HCC. In recent years, models have been demonstrated that could predict the response to immunotherapy [9], but more complex parameters need to be assessed in order to predict the survival of immunotherapy-treated HCC patients, taking into consideration the complex molecular profile of HCC [10-12]. Another score that has proved valuable in oncology is the neutrophil-to-lymphocyte ratio (NLR) [13,14]. Low NLR values have been associated with objective response and better survival in patients with solid tumors [15-18], and there is a rationale for using it in HCC, as it is a tumor with a complex microenvironment [19-21]. As a result, NLR has been previously correlated in some HCC cohorts with overall survival and response to atezolizumab–bevacizumab therapy [22-24].

Another important variable in HCC therapeutic management is the combined use of surgery or locoregional therapies with immunotherapy. Recently, the IMBRAVE050 study failed to show better survival rates for high-risk HCC patients who received adjuvant atezolizumab–bevacizumab post-resection, compared to active surveillance [25]. However, the administration of locoregional treatment before the initiation of immunotherapy can reshape the tumor microenvironment, making it less immune-resistant and increasing the chances of demonstrating an objective response with immunotherapy, and probably better survival [26,27].

The majority of HCCs used to arise on a substrate of liver cirrhosis [28]. However, the HCC landscape has changed dramatically over the last years. The increasing and more frequent use of antiviral drugs against hepatitis B and C viruses (HBV and HCV), in addition to the growing prevalence of steatotic liver disease, has led to more HCCs appearing in non-cirrhotic and non-viral patients, something that could definitely have a great impact on survival [29,30].

The aim of our study was to evaluate the efficacy of atezolizumab–bevacizumab combination therapy in patients with locally advanced and/or metastatic HCC not amenable to surgery or locoregional treatment, in relation to the baseline NLR values of histologically proved cirrhotic and non-cirrhotic HCC patients.

Results

Of the 57 patients, 45 were males (78.9%). Patients had a mean age of 66.3±11.3 years, a mean BMI of 27.23±4.8 kg/m², 19 had diabetes (33.3%), 33 viral etiology (57.9%), 29 were ALBI-I (50.8%), 22 had MVI (38.6%), 23 had EHD (40.4%), while 41 were categorized as BCLC-C (71.9%) and 16 as BCLC-B (28.1%) on study entry. In addition, 22 had AFP levels above 400 ng/mL (38.6%), 27 were categorized as proliferative subgroup HCCs (47.4%), 30 as non-proliferative (52.6%) and 31 had received prior HCC therapy (54.4%). In total, 14 patients (24.6%) underwent resection; only 4 were cirrhotic (28.6%) while the remaining 10 resected patients (71.4%) were non-cirrhotic. All resected patients in our study experienced early relapse, defined as HCC recurrence within 2 years post-resection. Locoregional treatment before immunotherapy onset was recorded in a total of 17 patients (29.8%), of whom 8 (47.1%) were cirrhotic, and 9 (52.9%) non-cirrhotic.

We performed our analysis in 3 steps. First of all, we divided the total of 57 HCC patients into 4 different groups, as described above. We then studied these groups according to their baseline characteristics. Taking into consideration that CP and MELD-Na scores were only used in patients with cirrhosis, and that none of the non-cirrhotic patients had signs of portal hypertension, we did not include CP stage and score, MELD-Na score or the presence of varices as baseline characteristics in this first analysis. On the other hand, we used the ALBI grade—which has already been widely used in patients with HCC and not exclusively in cirrhotic patients—as a baseline characteristic to measure hepatic reserve in all teams. The baseline characteristics of the 4 groups are presented in Table 1: the 4 groups were comparable in terms of all the baseline characteristics except for liver disease etiology (P<0.001), ALBI grade (P=0.014), MVI (P<0.001), BCLC stage at treatment initiation (P=0.014) and prior therapy (P=0.009). More specifically, non-cirrhotic patients tended to have non-viral disease, lower ALBI grade, rarely MVI, frequently BCLC-B stage and had often received prior treatment. Median OS (Fig. 1A) was 30, 10, 12 and 5 months for groups A, B, C and D, respectively and the log-rank test showed a statically significant result (P<0.001). We then performed Cox regression analysis for OS, in order to compare the survival of group D to the other 3 groups, after adjusting for all the baseline parameters for which the 4 groups were not comparable (etiology, ALBI, MVI, BCLC, prior therapy). According to the Bonferroni correction, the adjusted alpha level for statistical significance while comparing group D to groups A, B and C was 0.0167 (0.05/3) and we observed a statistically significant difference in OS when comparing group A to group D (HR 0.1, 95%CI 0.020-0.504; P=0.005), but not when comparing group C to group D (P=0.032) or group B to D (P=0.055), while the other covariates failed to achieve statistically significant P-values (Supplementary Fig. 1A). Similarly, median PFS (Fig. 1B) differed significantly (P<0.001) between the 4 groups (14, 4, 8 and 2 months for groups A, B, C and D, respectively). Cox regression for PFS, using the Bonferroni correction, showed a statistically significant difference when comparing group A to D (HR 0.140, 95%CI 0.043-0.459; P=0.006), whereas no statistical significance was found when comparing group B to D (P=0.070) and C to D (P=0.054), and no statistically significant P-values were found for the other covariates (Supplementary Fig. 1B).

Table 1 Baseline characteristics of the 4 groups created according to presence/absence of cirrhosis and baseline NLR status

Figure 1 Overall survival (A) and progression-free survival (B) for each of the 4 groups according to cirrhosis presence/absence and baseline NLR status

OS, overall survival; PFS, progression-free survival; NLR-H, high NLR group; NLR-L, low NLR group

We then conducted a Cox regression analysis for all patients, without categorizing them into groups, in order to assess the impact of cirrhosis and NLR separately for OS and PFS. We found that both presence of cirrhosis (HR 3.173, 95%CI 1.083-9.301; P=0.035) and high NLR (HR 3.076, 95%CI 1.337-7.078; P=0.008) were significantly associated with worse OS, independently of other potential confounders. For PFS, only a high NLR emerged as a statistically significant factor associated with worse PFS (HR 2.54, 95%CI 1.188-5.429; P=0.016), independently of all other confounders (Supplementary Fig. 2A,B).

The second analysis included only the cirrhotic HCC patients of our study (N=30). We divided those patients into 2 groups according to their baseline NLR status, and compared NLR-L and NLR-H patients according to their baseline characteristics (Supplementary Table 1). Those 2 groups were comparable for all baseline parameters, except for CP stage (P=0.021) and number (P=0.007). Median OS was significantly higher (P=0.002) in the cirrhotic NLR-L group (12 months) compared to NRL-H cirrhotic patients (5 months), and similarly PFS was significantly higher (8 months) in cirrhotic NLR-L compared to NLR-H patients (2 months, P=0.028; Fig. 2A,B). Cox regression analysis for OS showed that NLR was significantly associated with OS (HR 3.480, 95%CI 1.274-9.508; P=0.015), independently of CP stage and score, while no statistically significant results were achieved with regard to PFS (Supplementary Fig. 3A,B). We then compared cirrhotic patients according to prior therapy status for HCC: the 2 groups were comparable for all the baseline parameters evaluated (Supplementary Table 2) and did not show differences in median OS (P=0.201) or PFS (P=0.792) (Fig. 2C,D).

Figure 2 Overall survival (A) and progression-free survival (B) in cirrhotic HCC patients according to baseline NLR status. Overall survival (C) and progression-free survival (D) in cirrhotic HCC patients according to prior therapy status

OS, overall survival; PFS, progression-free survival; NLR-H, high NLR group; NLR-L, low NLR group

Finally, we compared the non-cirrhotic HCC patients (N=27) according to their baseline NLR status. The 2 groups were comparable for most of their baseline characteristics (Supplementary Table 3), except for prior therapy (P=0.016). Non-cirrhotic NLR-L patients showed a significantly higher median OS (30 months) compared to NLR-H non-cirrhotic patients (10 months, P=0.006) (Fig. 3A). Cox regression, adjusting for NLR status and prior therapy type, showed that only the latter factor and not NLR was significantly associated with OS (Supplementary Fig. 4A). Regarding PFS, non-cirrhotic NLR-L patients showed significantly higher PFS (15 months) than non-cirrhotic NLR-H patients (4 months, P=0.01, Fig. 3B). Cox regression analysis for PFS, adjusting for NLR and prior therapy type, showed that only prior treatment was significantly associated with PFS (Supplementary Fig. 4B). In our final analysis, we separated our non-cirrhotic patients in 2 groups according to prior HCC therapy status. Supplementary Table 4 shows the baseline characteristics of the 2 groups, which were comparable in terms of most factors, except for BCLC stage on treatment onset (P=0.03) and baseline NLR status (P=0.017). Those 2 groups differed significantly in terms of OS (30 months in patients with prior therapy vs. 8 months in patients without prior therapy, P<0.001) and PFS (24 months vs. 4 months, respectively, P<0.001, Fig. 3C,D). Cox regression analysis showed that only prior therapy, and not NLR or BCLC status, was significantly associated with OS in non-cirrhotic patients, and the same results also applied to PFS (Supplementary Fig. 5A,B). Lastly, we conducted Cox regression Kaplan–Meier analysis in order to present survival results (for both OS and PFS) at the mean of covariates, both for cirrhotic patients (using NLR as grouping variable) and for non-cirrhotic patients (using NLR and prior therapy separately as grouping variables). Results are presented in Supplementary Fig. 6,7.

Figure 3 Overall survival (A) and progression-free survival (B) in non-cirrhotic patients according to baseline NLR status. Overall survival (C) and progression-free survival (D) in non-cirrhotic HCC patients according to prior therapy status

OS, overall survival; PFS, progression-free survival; NLR-H, high NLR group; NLR-L, low NLR group

Discussion

It is noticeable that the majority of randomized trials, as well as real-world data in the HCC management field, do not separate cirrhotic from non-cirrhotic patients in their analysis of advanced HCC, despite the major impact of liver disease severity on the final clinical outcome [8,35,36]. Furthermore, non-cirrhotic patients in most of these studies are wrongly categorized as CP-A, considering that CP only applies to cirrhotic patients. Therefore, clinical studies of HCC should carefully distinguish between cirrhotics and non-cirrhotics when recruiting HCC patients, as survival outcomes may be affected by the presence of cirrhosis. In this study, we used easily obtained biomarkers, such as pre-treatment NLR, and recorded medical data, such as prior HCC treatment, in order to find possible correlations with survival in cirrhotic/non-cirrhotic patients.

In our study, we evaluated the efficacy of atezolizumab–bevacizumab in terms of OS and PFS, in a mixed population of cirrhotic and non-cirrhotic patients with unresectable HCC, according to baseline NLR. As expected, both OS and PFS were worse in cirrhotic compared to non-cirrhotic patients, and cirrhotic patients with high NLR values exhibited the worst outcomes (median OS 5 months, median PFS 2 months), irrespective of the other baseline parameters that could influence the survival outcome. Cirrhosis and high NLR values have been related to worse survival outcomes in untreated as well as treated HCC patients with all treatment modalities [22,37], including atezolizumab–bevacizumab therapy [23,24]. Both reports concerning the prognostic impact of NLR in patients treated with atezolizumab–bevacizumab [23,24] referred to fewer than 50 treated patients, without any histological definition of cirrhosis in most of them, while only 5-12.5% of patients were CP-B stage, and the studies had relatively short observation periods. In our study, high NLR values (>3) were an independent predictor of OS but not PFS, among histologically proven cirrhotic patients with HCC (23.3% in CP-B stage) treated with atezolizumab–bevacizumab. Several reports have highlighted the lack of a strong correlation between tumor-based endpoints, such as objective responses, and/or PFS and OS, especially among cirrhotic patients with advanced HCC under systemic treatment [38-40]. It has been suggested that the surrogacy of PFS with OS is variable, and mainly dependent on systemic treatment class (TKIs or immunotherapy-based) and evaluation treatment point [40].

In our study, we found that cirrhosis was a statistically significant factor for worse OS, as anticipated, along with high NLR values, heralding the importance of the separate evaluation of NLR status in cirrhotic and non-cirrhotic patients. As for PFS, in the total population of patients, only high NLR was independently associated with worse PFS; this suggests that NLR could be a very useful tool for predicting tumor response to immunotherapy, independently of cirrhosis and other confounding factors. Median OS was 12 months in cirrhotic patients with low baseline NLR, compared to only 5 months in those with high baseline NLR, a finding that was statistically significant. It is important to note that previous locoregional therapy (n=8) or resection (n=4) did not affect the survival rates of cirrhotic HCC patients (5-12 months for OS and 2-8 months for PFS).

NLR was not found to be an independent predictor of OS and PFS among non-cirrhotic HCC patients treated with atezolizumab–bevacizumab therapy. It is very interesting to note that 10 of the 27 non-cirrhotic HCC patients in our study population had been previously resected, and early disease recurrence was observed within 2 years from the time of resection, whereas locoregional therapy with incomplete response presented in 9 of 27 patients, prior to the initiation of atezolizumab–bevacizumab therapy. In the multivariate analysis, prior therapy, either resection or locoregional, was an independent predictor of long-term OS and PFS, among treated non-cirrhotic HCC patients. A beneficial reshaping of the tumor microenvironment by prior locoregional or surgical therapy has been suggested in many recently published studies, and trials evaluating combinations of these procedures with immunotherapy are already ongoing. Through this observational study, it could be supposed that prior therapy might possibly affect baseline NLR values prior to the start of immunotherapy, shifting the tumor microenvironment balance towards an immune-friendly environment before the initiation of immunotherapy, given that the majority of non-cirrhotic HCC patients (12/13, 92.3%) with a low baseline NLR had been previously treated with surgery or locoregional treatment, compared to only 50% (7/14, P=0.016) of those with high baseline NLR. This finding needs further investigation and evaluation in larger cohorts.

This study had some limitations. The small number of patients in each group could have led to biases, along with the retrospective analysis of the data. However, the grade of statistical significance of our values, in accordance with results from other studies [23,24], might suggest that NLR should be tested in cohorts of large randomized prospective trials to confirm its prognostic value in relation to survival. Another limitation is the small proportion of viral non-cirrhotic patients. This could be possibly explained by the increasing prevalence of steatotic liver disease in Europe, leading to the appearance of HCC in non-cirrhotic livers. On the other hand, this study was conducted in a single HCC referral center, where laboratory examinations were conducted by the same laboratory, and the same HCC specialists cooperated in multidisciplinary patient management, gathering of patient data and presentation of the results.