Association of Albumin-Bilirubin Grade and Sequential Treatment with Standard Systemic Therapies for Advanced Hepatocellular Carcinoma: A Retrospective Cohort Study Using a Japanese Administrative Database

Hepatocellular carcinoma (HCC) is the most common form of liver cancer and is a leading cause of death worldwide [1, 2]. Early-stage HCC can often be successfully treated by surgical resection, transplantation, or radiofrequency or microwave ablation. At later stages, treatment includes transarterial chemoembolization (TACE) for intermediate-stage HCC and systemic treatments for advanced HCC. Current Japanese treatment guidelines recommend sorafenib (approved in 2009) and lenvatinib (approved in 2018) as standard first-line treatment, and regorafenib (approved in 2017) as standard second-line treatment, for advanced HCC [3]. The choice of treatment depends not only on the cancer stage, but also on the extent of liver damage [2]. Most patients with HCC also have cirrhosis caused by pre-existing medical conditions, such as hepatitis or alcoholic liver disease. Because some treatments can further deteriorate liver function, the level of residual function should be considered when choosing treatment options and monitored during cancer treatment.

The Child–Pugh grading system has been used extensively to assess liver function [4]. However, the Child–Pugh system includes measures that are primarily subjective (ascites and encephalopathy) or are interrelated (e.g., ascites and serum albumin level), and may not be optimal for assessing liver function in patients with HCC to identify potential candidates eligible for systemic treatment according to current guidelines [5]. Recently, the albumin-bilirubin (ALBI) score has been proposed as an objective and simpler alternative to the Child–Pugh grading system for assessing liver function in patients with HCC [5, 6]. The ALBI scoring system is based only on objective laboratory measurements of serum albumin and total bilirubin, as determined by the equation: ALBI score = (log₁₀ bilirubin [μmol/L] × 0.66) + (albumin [g/L] × − 0.085). This is advantageous when used in retrospective database studies because laboratory measurement results are relatively more accessible compared with other clinical information in secondary data sources in Japan [7‐9]. Another advantage is that the ALBI score can be evaluated as a continuous measure, and can also be further divided into defined categories based on clinically relevant criteria. Evidence suggests that the ALBI score is a better predictor of prognosis following curative or palliative treatment than the Child–Pugh grade [6, 10‐12]. Although several studies from academic institutions such as university hospitals have been published recently [11, 13‐16], information about the distribution of ALBI scores/grades and their relationship with successful sequential systemic treatment in real-world clinical practice in Japan is still limited. Given that the current standard-of-care (SoC) systemic therapies were developed primarily in patients with better liver function (Child–Pugh grade A), it is important to assess whether the variation in liver function seen in clinical practice affects sequential treatment.

Administrative claims databases, which routinely collect information on daily medical activities, are considered a suitable and valuable data source for the investigation of real-world clinical practice. In this study, we used a hospital-based administrative database in Japan to investigate the distribution of ALBI scores and grades at baseline and at the end of first-line treatment in patients with HCC who were treated with SoC systemic therapy. We also described patient characteristics, the pattern of SoC treatment, and the relationship between ALBI score/grade and sequential SoC treatment, including duration of treatment.

The results of this observational cohort study using an administrative claims database indicate that baseline liver function, as assessed by ALBI score/grade, is associated with longer overall SoC treatment duration in real-world clinical practice in Japan. In addition, greater residual liver function at the end of index line treatment is associated with higher rates of transition to subsequent line SoC systemic therapy. These results suggest the importance of starting systemic therapy before liver function deteriorates and that efforts should be made to maintain liver function during systemic treatment to optimize patients’ ability to continue treatment. The rate of transition to subsequent line treatment was low, even after additional SoC options became available (i.e., in the lenvatinib and sorafenib B cohorts; Table 2). This was true also for those with greater residual liver function at the end of index line treatment (Table 5), indicating that other measures are needed to implement more successful sequential therapy in addition to maintaining liver function.

In this study, overall SoC treatment duration was significantly different across the four baseline ALBI grade groups. Median duration of treatment increased with lower ALBI grade and was shorter in patients with baseline ALBI grade 2b or 3 than in those with grade 1 or 2a, indicating that patients with better liver function are more likely to continue systemic therapy. Cox regression analysis confirmed that ALBI grade 2b and grade 3, as well as a higher Fibrosis-4 Index, were associated with shorter overall treatment duration, indicating that both liver function reserve and liver fibrosis affect prognosis. Similarly, the multinational GIDEON registry study reported that the duration of sorafenib treatment and overall survival were longer, and discontinuation because of adverse events was lower, in patients with Child–Pugh grade A than in those with Child–Pugh grade B [23]. Similar results were reported for the Japanese subset of the GIDEON study, whose patients closely resembled those of the current study in terms of age, sex, etiology of HCC, TNM cancer stage, and history of TACE before first-line treatment [24]. Several studies conducted in academic institutions in Japan have also demonstrated that baseline ALBI grade is a significant predictor of overall survival, as well as total duration of treatment [11, 13‐16]. A study by Ueshima et al. reported that baseline ALBI grade was associated with the tumor response rate, as well as the rate of discontinuation because of adverse events [11]. In addition, several studies have reported an association between ALBI grade and overall survival regardless of tumor stage [22, 25, 26]. Because deaths are captured in the MDV database only if they occur during hospitalization, we have not evaluated overall survival in this study. Nevertheless, overall treatment duration with systemic therapies is reported to be associated with overall survival in patients with unresectable HCC [14]. Collectively, these studies, together with our results, indicate that better liver function at the initiation of systemic therapy is associated with longer treatment and better outcomes. As suggested by previous studies, to optimize the likelihood of receiving subsequent treatment lines and to improve survival, systemic treatment should ideally be initiated while patients are still at ALBI grade 1 or at least 2a [13‐15]. Finally, although our study focused on current tyrosine kinase inhibitor SoC therapies, recent studies indicate that ALBI grade can also predict survival in patients treated with newer immuno-oncology drugs [27, 28], suggesting that preserving liver function will continue to be critical for successful sequential treatment as immuno-oncology drugs become more common.

Among patients who did not receive TACE during index line treatment, the transition rate to subsequent line treatment was higher in the sorafenib B cohort than in the sorafenib A cohort, as might be expected given the availability of second-line options after regorafenib was approved. For most patients in the sorafenib A cohort, no other SoC treatment options were available, and patients had to restart sorafenib as a subsequent line of therapy. The transition rate was also higher in the sorafenib B cohort compared with the lenvatinib cohort; this finding is consistent with previous Japanese studies suggesting that eligibility for second-line treatment was higher in patients treated with sorafenib (57.3% [29]) than in patients treated with lenvatinib (43.8% [15]). This difference in transition rate may partly be explained by a greater proportion of patients in the lenvatinib cohort (38.5%) with ALBI grade 3 at the end of index line treatment compared with the sorafenib cohorts (30.4% in sorafenib A, 23.5% in sorafenib B) [Fig. 2 of the ESM], given the observed association between ALBI grade at the end of the index line and the transition rate to the subsequent line (Table 5). Although it is not clear why the lenvatinib cohort had more ALBI grade 3 patients at the end of the index line, this finding potentially could be related to the longer treatment duration with lenvatinib than sorafenib that was seen in this study (data not shown) and reported in the randomized phase III REFLECT clinical trial (median 5.7 months for lenvatinib, 3.7 months for sorafenib [30]). Our study also found that patients with ALBI grade 1 or 2a at the end of index line treatment proceeded to subsequent line treatment more often than patients with ALBI grade 2b or 3, which may reflect the loss of patients with poorer liver function moving to palliative care or dying. Although these results are based on a relatively small number of patients, particularly with ALBI grade 1 or 2a, the findings suggest that maintenance of liver function during treatment may also improve patients’ ability to continue treatment. Our findings are consistent with those of Pinato et al., who reported that ALBI grade at the end of first-line sorafenib treatment was a predictor of post-sorafenib survival and could be used to guide the selection of patients most likely to benefit from second-line treatment [31]. A retrospective study from Taiwan also reported that more patients with ALBI grade 1 at the end of sorafenib treatment received subsequent therapies compared with patients with ALBI grade 2 (43.9% vs 28%) [32].

This study also documented the deterioration in liver function during index line treatment, even in the absence of TACE during the index line treatment, as shown by the change in ALBI score and grade. Between baseline and the end of index line treatment, mean ALBI score increased (worsened) by approximately 0.5 regardless of baseline ALBI grade. The ALBI grade also worsened during treatment; for example, the percentage of patients with ALBI grade 3 increased from 6.3% to 30.2% through the course of the index line treatment, whereas the percentage of patients with ALBI grade 1 decreased from 29.3% to 11.3%. These results are consistent with a retrospective study of patients in Taiwan in which the ALBI grade worsened in 54.1% of patients at the end of sorafenib therapy [32]. A more recent study conducted in Japan also reported significant ALBI score worsening from baseline to the end of sorafenib treatment (mean score − 2.42 vs − 2.02, p < 0.01), resulting in many patients being ineligible for second-line regorafenib [33]. Furthermore, a baseline ALBI score < − 2.53 (which roughly corresponds to ALBI grade 1/2a) was a significant predictor of regorafenib treatment initiation (p < 0.001) [33]. Deterioration of liver function during first-line lenvatinib treatment has also been reported [34]. However, these studies were not placebo controlled and thus a decline in liver function due to disease progression cannot be excluded. Our study also found that the ALBI grade was less likely to worsen during index line treatment in patients treated at designated cancer hospitals (Table 4), which may indicate the importance of consolidating care for patients with advanced HCC to experienced specialist hospitals.

The use of an administrative database covering approximately 22% of DPC hospitals in Japan allowed us to analyze a large number of patients with HCC being treated in real-world clinical practice, including from non-academic hospitals that are not usually involved in clinical research. The database also allowed us to understand changes in treatment sequences on a large scale, before and after post-sorafenib treatment options became available. The ability to assess liver function using only laboratory values is one of the advantages of the ALBI score compared with the Child–Pugh score, making it possible for us to evaluate liver function using the MDV database, in which clinical information necessary for the Child–Pugh score (presence and severity of ascites, hepatic encephalopathy) is not available. However, in the MDV database, laboratory test results were only available from hospitals that agreed to the use of test results for secondary research purposes. This limitation meant that baseline ALBI results were available for less than 10% of patients in each cohort, leading to smaller sample sizes for some analyses. Although this may have led to some selection bias in the analyzed population, we did not find any notable differences in available baseline characteristics between the full cohorts and the populations with baseline ALBI results. Another limitation is that only available variables were included in the regression models; therefore, confounding by unmeasured variables (e.g., performance status, tumor size, number of tumors, portal vein invasion) could not be excluded. Additional limitations of the MDV database include: (1) patients from smaller hospitals and clinics that have not adopted the DPC system are not included; (2) patients cannot be tracked beyond a single hospital; therefore, treatments outside that hospital cannot be captured, and patients who attend more than one hospital in the MDV database may be counted twice; and (3) certain information is not available, including reasons for treatment discontinuation (e.g., adverse events, tumor progression, liver dysfunction, death), tumor-related factors (e.g., tumor grade), deaths occurring outside of hospitalization, information on participation in outpatient-only clinical trials (information is available only when the trial requires hospitalization), and therapy line (therefore, index line treatment was assumed to be first-line treatment). In addition, some clinical information, such as body mass index and cancer stage, was available only from hospitalization discharge summaries and was therefore limited to patients who were hospitalized. Finally, although the transition rate to the subsequent line therapy in the lenvatinib cohort was low (Table 2), this may be biased by the shorter timeframe of lenvatinib availability (approximately 15 months from approval in March 2018 to the end of June 2019) compared with the sorafenib cohorts (approximately 24 months), resulting in a larger fraction of the lenvatinib cohort (approximately 50%) being excluded from the transition analysis, many of whom are likely to still be on index line treatment.

The results of this study using an administrative claims database suggest that better liver function reserve, as reflected by the ALBI score, at both the start and end of index line treatment is associated with successful sequential therapy in real-world clinical practice in Japan. Furthermore, despite the recent introduction of alternative systemic therapies, sequential treatment is not successfully implemented for many patients in real-world clinical practice in Japan. Because deterioration of liver function has a negative impact on sequential therapy, which in turn can adversely affect survival and quality of life, physicians should implement strategies to start systemic therapy while the patient still has good liver function and to maintain liver function as much as possible during systemic therapy.