Association between Adverse Events and Prognosis in Patients with Hepatocellular Carcinoma Treated with Atezolizumab Plus Bevacizumab: A Multicenter Retrospective Study

Simple Summary This study aimed to evaluate the correlation between adverse events (AEs) and overall survival (OS) in patients with unresectable hepatocellular carcinoma (HCC) treated with atezolizumab plus bevacizumab (atezo/beva). Liver injuries were significantly correlated with shorter survival. In a logistic regression analysis, fatigue ≥ grade 2, liver injury ≥ grade 3, and modified albumin–bilirubin grade 2b were identified as independent factors for discontinuation due to AEs. We concluded that the development of liver injury was a negative factor for OS and that we should be vigilant in monitoring AE during atezo/beva treatments. Abstract This study aimed to evaluate the correlation between adverse events (AEs) and overall survival (OS) in patients with unresectable hepatocellular carcinoma treated with atezolizumab plus bevacizumab (atezo/beva). This was a multicenter study in which 130 patients were enrolled. Hypertension and skin disorders had a significant correlation with longer survival (median survival time (MST): not reached vs. 14.3 months and not reached vs. 14.8 months, p = 0.001 and p = 0.047, respectively). In contrast, liver injuries were significantly correlated with shorter survival (MST: 14.7 months vs. not reached, p = 0.036), and the median development time was 21 days. In a logistic regression analysis, fatigue ≥ grade 2, liver injury ≥ grade 3, and modified albumin–bilirubin grade 2b were identified as independent factors for discontinuation due to AEs. The OS in the no discontinuation due to AE group was significantly longer than that in the discontinuation due to AEs group (MST not reached vs. 11.2 months, p = 0.001). We concluded that the development of liver injury was a negative factor for OS and that we should be vigilant in monitoring AE during atezo/beva treatments.


Introduction
Hepatocellular carcinoma (HCC) is the most common primary liver cancer and a leading cause of cancer-related deaths worldwide [1]. The reasons for these phenomena are that HCC is still often detected in the advanced stage [2] and that sorafenib (SORA) was the only approved systemic treatment [3] until a few years ago. However, a systemic treatment for HCC developed remarkably in the last few years, and various molecular

Patient Characteristics
The characteristics of the 130 enrolled patients are shown in Table 1

Initial and Best Therapeutic Outcomes of Atezo/Beva
The distribution of therapeutic responses to atezo/beva is presented in Table 2. In the initial RECIST evaluation, a complete response (CR) was observed in none of the patients, a partial response (PR) was observed in 35 patients (26.9%), stable disease (SD) was observed in 78 patients (60.0%), and progressive disease (PD) was observed in 17 patients (13.1%). The overall objective response rate (ORR) and disease control rates (DCR) were 26.9% and 86.9%, respectively. However, the overall ORR and DCR were 32.3% and 86.9%, respectively, for the best response.

Survival Analysis according to Each AE Profile following Atezo/Beva Treatment
The survival curves for each AE profile are shown in Figures 4 and S2. Patients wh developed liver injury had a significantly shorter survival time than those who did no (MST:14.7 months vs. not reached, p = 0.036) ( Figure 4A). In contrast, patients who deve oped hypertension or skin disorder survived significantly longer than those who did no (MST: not reached vs. 14.3 months and not reached vs. 14.8 months, p = 0.001 and p = 0.047 respectively) ( Figure 4B,C). Proteinuria, fatigue, appetite loss, and fever did not correlat with survival ( Figure S2A-D).

Survival Analysis According to Each AE Profile following Atezo/Beva Treatment
The survival curves for each AE profile are shown in Figures Figure 4 and S2. Patients who developed liver injury had a significantly shorter survival time than those who did not (MST:14.7 months vs. not reached, p = 0.036) ( Figure 4A). In contrast, patients who developed hypertension or skin disorder survived significantly longer than those who did not (MST: not reached vs. 14.3 months and not reached vs. 14.8 months, p = 0.001 and p = 0.047, respectively) ( Figure 4B,C). Proteinuria, fatigue, appetite loss, and fever did not correlate with survival ( Figure S2A-D).

Survival Analysis According to Early and Late Onset AEs
We defined the early-onset of AEs that develop within 6 weeks and the late-onset of AEs that develop after 6 weeks. There were no differences in OS between the two groups (early-onset AEs:17.9 months vs. late-onset AEs: not reached, p = 0.658) ( Figure S3).

Univariate and Multivariate Analyses of Factors Associated with OS
The ALBI grade, liver injury, hypertension, skin disorders, and post-progression treatments were identified as independent factors for OS in multivariate analysis (Table 4).

Survival Analysis according to Early and Late Onset AEs
We defined the early-onset of AEs that develop within 6 weeks and the late-onset of AEs that develop after 6 weeks. There were no differences in OS between the two groups (early-onset AEs:17.9 months vs. late-onset AEs: not reached, p = 0.658) ( Figure S3).

Univariate and Multivariate Analyses of Factors Associated with OS
The ALBI grade, liver injury, hypertension, skin disorders, and post-progression treatments were identified as independent factors for OS in multivariate analysis (Table  4).

Changes in ALBI Score in Patients with or without Developed Liver Injury
The median ALBI score in patients without liver injury recovered to nearly baseline values at 6 weeks after the introduction of atezo/beva (−2.42 vs. −2.39, p = 0.29). However, the median ALBI score in patients with liver injury did not improve compared with that at baseline (−2.39 vs. −2.17, p = 0.03) ( Figure 5).

Changes in ALBI Score in Patients with or without Developed Live
The median ALBI score in patients without liver injury recov values at 6 weeks after the introduction of atezo/beva (−2.42 vs. −2 the median ALBI score in patients with liver injury did not impro at baseline (−2.39 vs. −2.17, p = 0.03) ( Figure 5). Figure 5. Changes in ALBI score 3 and 6 weeks after the atezo/beva treatm there is liver injury. The median ALBI score in patients without liver i baseline values at 6 weeks after introducing atezo/beva (−2.42 vs. −2.39, p dian ALBI score in patients with liver disorder has not improved compare vs. −2.17, p = 0.03). HCC, hepatocellular carcinoma; ALBI, albumin-bili zumab plus bevacizumab.

Decision-Tree Analysis for the Discontinuation of Atezo/Beva due
In this study, the rate of discontinuation due to AEs in all su time of study cessation (Table S1). To determine the profiles assoc tion owing to AEs, a decision tree analysis was performed. Fatigu first splitting variable for the rate of discontinuation due to AEs. A continuation due to AEs was only 26.7% in patients with fatigue < continuation due to AEs was 80% in patients with fatigue ≥grad tigue <grade 2, the second splitting variable was liver injury. In p <grade 2 and liver injury ≥grade 3, the rates of discontinuation du 87.5%, respectively. In patients with liver injury <grade 2, the AL Figure 5. Changes in ALBI score 3 and 6 weeks after the atezo/beva treatment according to whether there is liver injury. The median ALBI score in patients without liver injury recovered to nearly baseline values at 6 weeks after introducing atezo/beva (−2.42 vs. −2.39, p = 0.29). However, the median ALBI score in patients with liver disorder has not improved compared to baseline values (−2.39 vs. −2.17, p = 0.03). HCC, hepatocellular carcinoma; ALBI, albumin-bilirubin; atezo/beva, atezolizumab plus bevacizumab.

Decision-Tree Analysis for the Discontinuation of Atezo/Beva Due to AEs
In this study, the rate of discontinuation due to AEs in all subjects was 32.2% at the time of study cessation (Table S1). To determine the profiles associated with discontinuation owing to AEs, a decision tree analysis was performed. Fatigue was identified as the first splitting variable for the rate of discontinuation due to AEs. Although the rate of discontinuation due to AEs was only 26.7% in patients with fatigue <grade 2, the rate of discontinuation due to AEs was 80% in patients with fatigue ≥grade 2. In patients with fatigue <grade 2, the second splitting variable was liver injury. In patients with liver injury <grade 2 and liver injury ≥grade 3, the rates of discontinuation due to AEs were 20.5 and 87.5%, respectively. In patients with liver injury <grade 2, the ALBI grade was identified as the next level of splitting variables. In patients with fatigue <grade 2 concomitant with liver injury <grade 2 and ALBI grade 1 or 2a, the discontinuation rate due to severe AEs was only 12.8% ( Figure 6). R REVIEW 9 of 13 Figure 6. Profiles associated with discontinuation due to AEs in patients with HCC treated with atezo/beva. Decision-tree algorithm for discontinuation due to AEs. The pie graphs indicate the percentage of no discontinuation due to AEs (white)/discontinuation due to AEs (black) in each group. HCC, hepatocellular carcinoma; atezo/beva, atezolizumab plus bevacizumab.

Logistic Regression Analysis for Discontinuation due to AEs
Fatigue ≥ grade 2, liver injury ≥ grade 3, and m-ALBI grade 2b were selected as variables in a stepwise logistic regression analysis. In the logistic regression analysis, all three variables were identified as independent factors for discontinuation due to AEs (Table 5).

Additional Treatments after the Discontinuation of Atezo/Beva
Until the time of study cessation, 96 (73.8%) patients had their atezo/beva treatment discontinued, and 68 (70.8%) patients received subsequent treatments (Table 6). Among these patients, 24 patients (35.2%) were treated with LEN, and 9 patients (13.2%) were treated with ramucirumab or transcatheter arterial chemoembolization. Profiles associated with discontinuation due to AEs in patients with HCC treated with atezo/beva. Decision-tree algorithm for discontinuation due to AEs. The pie graphs indicate the percentage of no discontinuation due to AEs (white)/discontinuation due to AEs (black) in each group. HCC, hepatocellular carcinoma; atezo/beva, atezolizumab plus bevacizumab.

Logistic Regression Analysis for Discontinuation Due to AEs
Fatigue ≥ grade 2, liver injury ≥ grade 3, and m-ALBI grade 2b were selected as variables in a stepwise logistic regression analysis. In the logistic regression analysis, all three variables were identified as independent factors for discontinuation due to AEs (Table 5).

Discussion
This study showed that liver injury, hypertension, and skin disorders were important factors in predicting the survival of patients with HCC treated with atezo/beva. Moreover, we demonstrated that ≥grade 2 fatigue, ≥grade 3 liver injury, and m-ALBI grade 2b were independently associated with discontinuation due to AEs in patients with HCC treated with atezo/beva. This study demonstrated that the ORR and DCR in the included patients were 32.3% and 86.9%, respectively. Furthermore, incidence rates of any grade of and grade ≥3 AEs were 96.9% and 36.1%, respectively, in patients with HCC treated with atezo/beva. According to the IMbrave 150 trial, the ORR, DCR, and rate of grade ≥3 AEs were 30.0%, 74.0%, and 43.0% in patients with unresectable HCC treated with atezo/beva, respectively [9]. Thus, the therapeutic effect and incidence rate of AEs in the present study appear to be similar to those in previous reports, suggesting that the enrolled subjects, treatment effects, and use of atezo/beva in our study are standard.
In this study, patients who developed liver injury during atezo/beva treatments had a significantly shorter survival time than those who did not. Previous reports have shown that liver injuries are associated with poor prognosis in patients with cancer treated with ICI therapy [17][18][19]. Interestingly, the prognosis of patients with HCC with grade 1 or 2 liver injury during ICI treatments was also poor [20]. Chen et al. reported that patients with liver injury in liver cancers have the highest chance of concomitant hyperbilirubinemia and biliary obstruction as cholestasis may contribute to the development of liver injury [17]. In this study, the median ALBI score in patients with liver injury did not improve compared to that at baseline (−2.39 vs. −2.17, p = 0.03). Moreover, Mouri et al. reported that patients administered with systemic steroids had fewer therapeutic effects than those who did not [21]. In fact, 14% (8/57) of patients who developed liver injury received systemic steroid treatments in our study. This may have contributed to the poor prognosis associated with the appearance of liver injury throughout the atezo/beva treatment. In contrast, hypertension and skin disorders were better prognostic factors for survival. Several studies reported that hypertension is a clinical biomarker for responses to bevacizumab [22,23]. Moreover, we previously reported that the appearance of hypertension can be a clinically promising early surrogate marker for predicting the survival of patients treated with LEN [11]. Furthermore, one of the reasons is that patients who developed hypertension had a significantly higher DCR than those who did not in this study (Table S2). As for skin disorders, the development of immune-related skin disorders was found to be correlated with favorable outcomes and its practical feasibility as a potential predictive surrogate marker [24,25]. The reason why hypertension and skin disorders could be positive prognostic markers is unclear. However, to our knowledge, this study is the first to reveal that hypertension and skin disorders could be positive predictive markers for the prognosis of patients undergoing atezo/beva treatment. Thus, further studies are needed to clarify these reasons.
Recently, a previous study reported that OS correlated with the treatment duration of systemic therapy [26]. In this study, we found that treatment durations were significantly longer in patients who developed hypertension and skin disorders than in patients who did not (median treatment duration: 7.8 vs. 4.9 months and 8.5 vs. 5.3 months, p = 0.001 and p = 0.027, respectively). These AEs are relatively manageable using supportive drugs such as antihypertensive drugs and steroids. Therefore, as possible, it would be better to continue the treatment with the management for these AEs. However, if the effects of treatment are no longer confirmed, we should promptly switch to second-line treatments. Moreover, the treatment duration was significantly shorter in patients who developed liver injury than in patients who did not (median treatment duration: 4.3 vs. 6.7 months, p = 0.034). Additionally, the median ALBI score in patients with liver in-jury did not improve compared with that at baseline, which suggests that the development of liver injury induces the deterioration of liver function. Thus, AE management is needed for patients who developed liver injury as possible, but switching to second-line treatments should be also considered at the time of the development of liver injury, particularly severe liver injury In this study, fatigue ≥ grade 2, liver injury ≥ grade 3, and m-ALBI grade 2b were identified as independent factors for discontinuation due to AEs using multivariate analyses. Moreover, we also found that fatigue was the initial splitting variable for the rate of discontinuation due to AEs in patients with HCC treated with atezo/beva, followed by liver injury and m-ALBI grade in the decision tree analysis. Several previous studies have reported that fatigue, liver injury, and m-ALBI grade are predictive factors for therapeutic effects or OS in patients treated with systemic therapy [18,27,28]. Fatigue can be detrimental to the patients' quality of life [29], and preserved liver function is a favorable factor related to eligibility for post-treatment [30]. Sequential systemic therapy has recently been considered an effective strategy for the treatment of unresectable HCC. Furthermore, discontinuation due to AEs reduces the rate of transition to the next treatment and must be avoided. In our study, among the discontinuation of atezo/beva treatment, 68 patients (70.8%) received subsequent treatment. However, patients who had their treatment discontinued due to Aes that were significantly less likely to receive subsequent treatments than those who did not (38.7% vs. 86.1%, p < 0.001). Moreover, the median OS in the not discontinued due to AEs group was significantly longer than in the discontinued due to AEs group (MST: not reached vs. 11.2 months, p = 0.001, Figure S4), indicating that our results were in accordance with those of previous reports treated with systemic therapy [27,28,31]. Thus, clinicians should be vigilant in monitoring AE during atezo/beva treatment. Moreover, the establishment of a comprehensive grading system to predict discontinuation due to AEs is needed in the management of atezo/beva treatment.
This study has some limitations. First, this was a retrospective study. Second, although atezo/beva treatment was administered after the disappearance of the effect of previous treatment, we cannot deny that previous MTAs influenced the development of AEs due to atezo/beva. Third, we investigated post-treatment after atezo/beva treatment and did not obtain a sufficient observation period. Thus, further prospective validation studies with long-term follow-up are required.

Conclusions
In this study, we showed that fatigue, liver injury, and m-ALBI grade 2b were independently associated with discontinuation due to AEs in patients with HCC treated with atezo/beva. The study revealed that the types of developed AEs were important in predicting survival in atezo/beva treatments. The establishment of appropriate AE management is needed to further contribute to survival in patients with HCC.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers14174284/s1, Figure S1: Study design. A total of 153 patients with HCC were enrolled; Figure S2: Overall survival time in patients who developed each AE in atezo/beva treatment; Figure S3: Overall survival time in patients with HCC according to early and late onset AEs; Figure S4: Overall survival time in patients with HCC who did or did not require discontinuation due to AEs. Table S1: The reasons for treatment discontinuation in atezo/beva; Table S2: Therapeutic responses according to AEs (n = 130).