The effectiveness and safety of immune checkpoint inhibitors in conversion or neoadjuvant therapy for hepatocellular carcinoma: a meta-analysis and systematic review.

Introduction
Liver cancer demonstrates the third-highest incidence and sixth-highest mortality rates worldwide [1]. Among the various subtypes, hepatocellular carcinoma (HCC) stands as the most prevalent type of liver cancer. Typically, HCC is detected during the middle to advanced stages, with less than 30% of cases deemed operable at the early stages [2, 3]. Conversely, patients with inoperable HCC exhibit significantly lower survival rates compared to those who undergo surgery [2, 3]. As per the guidelines provided by the European Association for the Study of the Liver (EASL) and the American Association for the Study of Liver Diseases (AASLD), individuals with Barcelona Clinic Liver Cancer (BCLC) grade B or C are recommended to undergo transarterial chemoembolization (TACE) or systemic therapy [4, 5]. However, a meta-analysis revealed that patients with BCLC grade B or C who received surgical intervention experienced superior survival outcomes compared to those treated with TACE or sorafenib [6]. The finding strongly advocated for the inclusion of surgery as a recommended course of action, even for patients with BCLC grade B or C, provided that surgical options are available. The Chinese liver cancer guidelines similarly endorse surgical resection as the preferred treatment for select patients with BCLC beyond grade A [7].
Immune checkpoint inhibitors (ICIs) targets checkpoint molecules to reverse T-cell functional exhaustion, thereby reinstating anti-tumor immune responses [8]. Inhibition of the Vascular Endothelial Growth Factor (VEGF) pathway induces vascular normalization, which concurrently diminishes immunosuppressive cells while augmenting the activity of dendritic cells and effector T cells [8]. The advent and clinical implementation of tyrosine kinase inhibitors (TKIs) and ICIs have significantly enhanced the survival outcomes of patients diagnosed with unresectable middle to advanced liver cancer following systematic treatment. Additionally, a subset of patients experiences partial or complete tumor remission [9]. Consequently, certain cases of advanced HCC undergo downstaging, rendering them amenable to resection. This development has opened avenues for the application of neoadjuvant therapy or conversion therapy in the field of HCC [10, 11]. Therefore, scholars have embarked upon the study of HCC conversion therapy, revealing that patients with middle or advanced HCC who underwent surgical resection after conversion therapy exhibit improved survival rates compared to those who solely underwent surgical resection [12].
Currently, a multitude of TKIs, VEGF-antibody, and ICIs are employed for the management of unresectable HCC [13, 14]. In addition to the synergistic combination of ICIs and TKIs or atezolizumab with bevacizumab, certain researchers have demonstrated favorable outcomes by integrating TACE or hepatic arterial infusion chemotherapy (HAIC) with those medicine [15–17]. However, the efficacy of conversion or neoadjuvant therapy exhibits considerable variability across different studies and is contingent upon individual patient characteristic [15, 16, 18–35]. Our investigation primarily focused on examining the therapeutic role of immunotherapy in combination with targeted agents, anti-angiogenic medications, and/or localized treatment modalities such as TACE, HAIC, and stereotactic body radiotherapy (SBRT) in the context of HCC conversion or neoadjuvant therapy. The conversion rate, incidence of adverse events (AE), and tumor response were comprehensively summarized to furnish clinicians with an invaluable point of reference.

Methods
An ethics statement is not applicable because this study is based exclusively on published literature. This systematic review is registered in PROSPERO (registration no. CRD42023446501).

Search strategy
We searched PubMed, Embase, Web of Science, and Cochrane Library databases from inception to June 11, 2025. We searched the literature in PubMed by combining the following terms and keywords: “neoadjuvant”, “conversion”, “immune checkpoint inhibitors”, “hepatectomy”, and “hepatocellular carcinoma”, to retrieve relevant literature. The search was limited to English-language publications. Supplementary Table S1 provides detailed information regarding the search strategies employed across all databases. Additionally, references in the relevant studies were manually searched for potentially eligible studies.

Inclusion and exclusion criteria
The inclusion criteria for eligible studies encompassed investigations that examined the use of ICIs as monotherapy or in combination with TKIs, VEGF antibody, or local therapies such as TACE, HAIC, and radiotherapy as neoadjuvant or conversion therapy for HCC. Regard to studies focused on conversion therapy, the result of conversion rate to surgery is necessary. Regard to studies focused on neoadjuvant therapy, the result of resection rate is necessary. The study design included observational studies, non-comparative studies, comparative studies, and randomized controlled trials. Abstracts, case reports, reviews, and studies not published in English were excluded from consideration. In instances where multiple studies featured overlapping patient cohorts, preference was given to the most recent publication or the study with the largest sample size.

Outcomes of interest
For conversion therapy, the primary outcome was the conversion rate. Additional outcomes encompassed pathological complete response (pCR), complete response (CR) rate, partial response (PR) rate, objective response rate (ORR), disease control rate (DCR), and adverse events (AEs). These outcomes were evaluated utilizing either Response Evaluation Criteria In Solid Tumors (RECIST) or modified Response Evaluation Criteria In Solid Tumors (mRECIST). Pathological complete response (pCR) was defined as no viable tumor cells in the resection specimens. Major pathological response (MPR) defined as 90%–99% tumor necrosis in resected tissue. Adverse events (AE) resulting from conversion therapy regimens were also assessed, with grade 3 and above AEs being calculated accordingly.
For neoadjuvant therapy, the primary outcome was resection rate. Additional outcomes were similar to conversion therapy, including tumor response and AE related to neoadjuvant therapies. Overall survival (OS) and recurrence-free survival (RFS) in the studies compared the prognosis between patients receiving neoadjuvant therapy upfront surgery to those receiving surgery only were also extracted. The remaining pertinent details were extracted and meticulously recorded on the tables.

Definition and measures of outcomes
Conversion rate was defined as the percentage of patients who underwent liver resection following conversion therapy. Resection rate was defined as the percentage of patients who underwent liver resection following neoadjuvant therapy. OS was calculated from the date of liver resection to the date of death or the last follow-up. Recurrence-free survival (RFS) was defined as the time interval from the date of liver resection to tumor recurrence or death. The definitions of CR and PR can be found in the RECIST or mRECIST criteria. ORR was defined as the percentage of patients achieving CR or PR, while DCR represented the percentage of patients achieving CR, PR, or stable disease (SD) according to RECIST or mRECIST criteria. AEs were assessed using the Common Terminology Criteria for Adverse Events v4.0.

Quality assessment and data extraction
Two authors (LH and YX) independently evaluated the quality of each study, using the methodological index for non-randomized studies (MINORS) tool which comprised 12 items, with each item allocated a maximum score of 2 points. In evaluating the included studies, the first eight items of the MINORS tool were employed including the clarity of the study’s objectives, consistency in patient inclusion, collection of anticipated data, appropriateness of endpoints reflecting the study’s objectives, objectivity in endpoint evaluation, adequacy of follow-up, dropout rates below 5%, and the consideration of sample size estimation. Each item was scored on a scale from 0 to 2: 0 indicated unreported, 1 reported but inadequately, and 2 reported and fully detailed, with an ideal score of 16. Furthermore, two authors (LH and YQ) independently extracted the relevant data, including study details, patient characteristics, tumor characteristics, and information about conversion or neoadjuvant therapy regimens with predesigned and standardized forms. In the event of any disagreement, discussion, or a third opinion (RW) was sought to resolve.

Statistical analysis
The conversion and resection rate and its corresponding 95% confidence interval (CI) were calculated using the ‘meta’ package. The distribution of the sample rate was detected, and the original sample rate should conform to the normal distribution. The inverse variance method was used to determine the hazard ratio (HR) and 95% confidence interval (CI) values. Heterogeneity was assessed using the χ2 method, with I2 ≤ 25% indicating low heterogeneity, 25% < I2 ≤ 50% indicating moderate heterogeneity, and I2 > 50% indicating high heterogeneity. We used the random-effects model if I2 > 50%. A sensitivity analysis was conducted to assess the robustness of our conclusion. Funnel plot and Egger’s test was conducted to detect the publication bias. Subgroup analyses were performed based on different treatments. Meta regression was conducted to discover the source of heterogeneity. Statistical significance was determined using a two-tailed p-value threshold of 0.05. All statistical computations were conducted using the R language (version 4.3.1).

Results

Study search and selection
A comprehensive search yielded a total of 1310 records, of which 816 remained after removing duplicates. Subsequently, 687 papers were excluded after reviewing the title and abstract, leaving 129 articles for further consideration. Among these, 80 articles were further excluded for various reasons, as shown in Fig. 1. Consequently, 49 studies were selected for inclusion in our meta-analysis [17, 18, 22, 23, 25, 28, 32, 36–76], with 36 of these studies focusing on conversion therapy [17, 18, 22, 23, 25, 28, 32, 33, 36–46, 48, 49, 51, 52, 54, 55, 57, 58, 60–62, 64, 66, 67, 69–71]and the remaining eight examining neoadjuvant therapy [47, 50, 53, 56, 59, 63, 65, 68, 72–76].

Study characteristics
The 36 studies, which focused on conversion therapy, comprised seven phase 2 clinical trials, one prospective study, and 28 retrospective studies, involving a total of 3497 patients. These studies originated from Japan (n = 3) and China (n = 33); The thirteen studies, which focused on neoadjuvant therapy, consisted of two phaseⅠb trial, three phase Ⅱ trials, one prospective study, and seven retrospective studies, involving 569 patients. These studies were conducted in the USA (n = 4) and China (n = 9). Detailed characteristics of included studies and patients can be found in Tables 1 and 2, S2, and S3. The quality of the included studies was assessed by the MINORS tool, with two study scoring 16 points, eight scoring 12 points, and the remaining scoring 14 points (Table S4).
As indicated in Table S5, the regimens utilized in conversion therapy exhibited considerable variation. These included ICI combined with TKI or bevacizumab, Atelizumab and bevacizumab (A + T), HAIC + ICI + TKI or bevacizumab, TACE + ICI + TKI or bevacizumab, TACE + HAIC + ICI + TKI or bevacizumab. Notably, the ICIs encompassed pembrolizumab, nivolumab, camrelizumab, toripalimab, sintilimab, and tislelizumab; The TKIs employed included lenvatinib, sorafenib, and apatinib, with lenvatinib accounting for the largest proportion. The chemotherapy regimen used in HAIC included FOLFOX and RALOX, with the former being more prevalent. The details of TACE exhibited slight variations across studies. Table S6 provides a comprehensive overview of the regimens employed in neoadjuvant therapy, albeit with small sample sizes.

Outcomes for conversion therapy

Conversion rate
All included studies reported conversion rates, ranging from 2% to 69%. A meta-analysis with the random effects model revealed a pooled proportion of conversion rate of 0.23 (95% CI: 0.18–0.29) (Fig. 2).

Perioperative outcomes after hepatectomy
Five studies reported perioperative outcomes of patients received hepatecomy after successful conversion (Table S7).

Pathological complete response (pCR)
Fourteen studies assessed the pCR rate, and the pooled data indicated a pCR rate of 0.08 (95% CI: 0.05–0.12), as presented in Figure S1.

Tumor response evaluated by RECIST
Twenty-four studies evaluated the ORR, and pooled data revealed an ORR of 0.43 (95% CI: 0.36–0.49) (Figure S2). Additionally, twenty-four studies assessed the DCR with a pooled DCR of 0.85 (95% CI: 0.80–0.90) (Figure S3).

Tumor response evaluated by mRECIST
Thirty-four studies evaluated the ORR, and the pooled data demonstrated an ORR of 0.62 (95% CI: 0.56–0.67) (Figure S4). Furthermore, thirty-three studies evaluated the DCR, and the pooled data indicated a DCR of 0.87 (95% CI: 0.84–0.90) (Figure S5).

AE
Twenty-three studies provided data on all grade AEs that were included in the analysis. The pooled data revealed an all-grade AE rate of 0.95 (95% CI: 0.92–0.98) (Figure S6). While twenty-four studies reported grade 3 and higher AEs with a pooled ORR rate of 0.39 (95% CI: 0.32–0.46) (Figure S7). Adverse events that were reported in more than ten original studies were also analyzed. Details of these results were shown in Figure S8, S9, S10, S11, S12, and S13.

Sensitivity analysis, publication bias, meta regression, and subgroup analysis
The sensitivity analysis conducted for conversion rate indicated that the result was stable (Figure S14). Publication biases existed in most outcomes (Figure S15). Meta-regression analysis did not find the source of heterogeneity (Figure S16). Subgroup analyses were performed based on different treatments (Figure S17, S18 and S19). The conversion rate for the regimen combining ICI, TKIs or bevacizumab, and TACE was 0.32 (95% CI: 0.23–0.43), while the conversion rate for the regimen combining ICI, TKIs, and HAIC was 0.32 (95% CI: 0.20–0.47). Additionally, the conversion rate for the A + T regimen was 0.04 (95% CI: 0.02–0.06) (Figure S18). Figure S18 also provided subgroup analyses for other outcomes. Similarly, tumor response and adverse events were different among the patients who achieved different treatments. Figure S19 provided a more detailed illustration of the impact on various outcome measures in conversion therapy by combining different local treatments, molecular targeted therapies, and immune checkpoint inhibitors. The HAIC + Apatinib + Camrelizumab regimen achieved the highest conversion rate, followed by HAIC + Bevacizumab biosimilar + Sintilimab, HAIC/TACE + Bevacizumab + Sintilimab, and TACE + Lenvatinib + Camrelizumab. HAIC chemotherapy regimens (FOLFOX vs. RALOX) and TACE embolic agent types (conventional vs. drug-eluting beads) may also affect conversion outcomes. We did subgroup analyses based on HAIC chemotherapy regimens and TACE embolic agent types and showed in Figure S20, S21, S22, S23, S24, and S25. In most studies, conventional TACE and the FOLFOX regimen for HAIC were employed. Among TACE-based treatments, conventional TACE resulted in a higher conversion rate (Figure S20), whereas in TACE + HAIC-based treatments, drug-eluting beads TACE led to a higher conversion rate (Figure S22).The FOLFOX regimen is more commonly used in HAIC and demonstrates a higher conversion rate compared to the RALOX regimen (Figure S21 and S23). In the selection of molecular targeted agents, more centers opted for lenvatinib and bevacizumab (Figure S26). Figures S27 and S28 illustrate the outcomes of combining lenvatinib or bevacizumab with various local treatments and immunotherapies. Figure S29 presents the conversion rates and other outcome metrics achieved by different immunotherapy regimens.

Outcomes for neoadjuvant therapy

Resection rate
All included studies reported resection rates, ranging from 69% to 100%. A meta-analysis with the random effects model revealed a pooled proportion of conversion rate of 0.87 (95% CI: 0.85–0.90) (Fig. 3). Thirty-eight patients did not receive surgery due to tumor progression (Table S8).

OS and RFS
Out of the included studies, five reported HR of OS. The random-effect model was used because of high heterogeneity among studies. The pooled data showed that neoadjuvant therapy was not associated with a statistically significant improvement in OS (HR, 0.52; 95% CI, 0.25–1.06; p = 0.0714). (Figure S30). Five studies reported HR of RFS. The random-effect model was used because of high heterogeneity among studies. The pooled data showed that neoadjuvant therapy was not associated with a statistically significant improvement in RFS (HR, 0.6; 95% CI, 0.31–1.02; p = 0.0572) (Figure S31).

PCR and MPR
Eight studies evaluated the pCR, and pooled data revealed a pCR, of 0.13 (95% CI: 0.06–0.19) (Figure S32). Additionally, seven studies assessed the MPR with a pooled MPR of 0.3 (95% CI: 0.10–0.16) (Fig. 33).

Tumor response evaluated by RECIST
Seven studies evaluated the ORR, and pooled data revealed an ORR of 0.20 (95% CI: 0.08–0.32) (Figure S34). Additionally, seven studies assessed the DCR with a pooled DCR of 0.94 (95% CI: 0.90–0.99) (Figure S35).

Tumor response evaluated by mRECIST
Eight studies evaluated the ORR, and the pooled data demonstrated an ORR of 0.60 (95% CI: 0.41–0.79) (Figure S36). Furthermore, 4 studies evaluated the DCR, and the pooled data indicated a DCR of 0.95 (95% CI: 0.92–0.99) (Figure S37).

AE
Nine studies provided data on all grade AEs that were included in the analysis. The pooled data revealed an all-grade AE rate of 0.85 (95% CI: 0.7–0.96) (Figure S38). While eight studies reported grade 3 and higher AEs with a pooled rate of 0.22 (95% CI: 0.16–0.28) (Figure S39).

Sensitivity analysis, publication bias, and subgroup analysis
The sensitivity analysis conducted for resection rate indicated that the result was stable (Figure S40). Significant heterogeneity was detected in the analysis of resection rate (Figure S41). Subgroup analyses were performed based on different treatments. The resection rates for the regimen combining ICI, TKIs, and local treatment were all above 90% (Figure S42). Figure S42 also provided subgroup analyses for other outcomes. Patients who received ICI plus local therapy achieved higher rate of ORR.

Discussion
This meta-analysis examined the therapeutic efficacy and adverse effects of current immune checkpoint inhibitors in HCC conversion or neoadjuvant therapy. Subgroup analysis was conducted based on different treatment modalities, providing valuable insights for clinicians.
In terms of numerical results, the overall conversion rate was 23%. Notably, the combination of ICIs and TKIs with TACE or HAIC demonstrated higher conversion rates of 32% and 32%, respectively. Conversely, Atezolizumab plus Bevacizumab exhibited the lowest conversion rate of 4%. ICIs plus TKIs or bevacizumab combined with TACE or HAIC demonstrated the highest ORR. These findings suggested a strong correlation between conversion rate and ORR, where a higher ORR indicated a higher conversion rate [28, 77]. A previous meta-analysis focused on the efficacy and safety of the systemic conversion therapy including TKI, ICI, loco-regional therapy and the combination of them [78]. However, in our study, we focused on the efficacy and safety of ICI and the combination of ICI with loco-regional therapy and molecular target therapies. The similar finding between the two studies is that TKI + ICI + loco-regional therapy achieved relatively higher conversion rate.
In terms of safety, special emphasis was placed on grade ≥ 3 treatment-related AEs) which were classified as serious AEs. The analysis revealed that serious AEs occurred in 39% of cases, while all-grade AEs were observed in 94% of patients, indicating a high frequency of treatment-emergent toxicities associated with conversion therapies. These findings underscore the critical importance of implementing appropriate management strategies for treatment-related side effects. Among the 22 types of AEs reported in at least 10 studies, the most frequently documented toxicities included gastrointestinal reactions, hepatic dysfunction, myelosuppression, and hand-foot syndrome. These observations highlight the necessity for close monitoring and proactive management of these specific adverse effects during conversion therapy.
Our meta-analysis comprehensively reviewed the conversion rate and safety of various treatment modalities, including ICIs, VEGF antibody, HAIC, TACE, TKI, TKI combined with ICIs, and local therapy. We found that T + A achieved a lower conversion rate than the combination of ICIs with TKI and local therapy, suggesting that the combination of ICIs with TKI or local therapy can help to improve the conversion success rate. However, it is important to note that due to limited primary research data, subgroup analysis based on different combinations of ICIs, TKI, VEGF antibody, and local therapy could not be performed. In the era of immunotherapy and targeted therapy, the utilization of ICIs combined with TKIs or VEGF antibody or/and local therapy such as TACE and HAIC are increasing in the treatment of HCC. Therefore, it is imperative to summarize the conversion rate and safety outcomes associated with these treatment options in HCC conversion therapy. The strength of this study lies in its comprehensive synthesis of the latest research and its subgroup analysis, highlighting both the advantages and disadvantages of different treatment combinations. Based on these valuable insights, clinicians can make informed decisions and select the most suitable treatment options for their patients.
Regarding neoadjuvant therapy for HCC, multiple regimens have been proposed; however, the limited sample size in each regimen precluded the conduct of a meta-analysis. Nevertheless, the available survival outcomes from individual studies indicated the effectiveness of neoadjuvant therapy, with a relatively low proportion of patients being deemed ineligible for surgery due to tumor progression. Furthermore, the incidence of side effects was acceptable. Zhao et al. provided a summary of additional regimens outlined in conference abstracts [79]. Through a meta-analysis incorporating data from both published articles and conference abstracts, they concluded that neoadjuvant immune checkpoint inhibitors may offer therapeutic advantages in terms of histopathological response and manageable toxicity profiles in patients with resectable HCC. We eagerly anticipate further studies with larger sample sizes to validate this conclusion.
We further compared the oncological outcomes between HCC patients who underwent neoadjuvant therapy followed by surgery and those treated with surgery alone. The analysis revealed that the addition of neoadjuvant therapy ​did not yield a statistically significant improvement​ in OS or RFS. ​However, these findings should be interpreted with caution due to the limited number of included studies and substantial tumor heterogeneity among the analyzed patient populations. Of the five original studies [47, 50, 53, 56, 63] included in the meta-analysis that compared prognostic outcomes between neoadjuvant therapy followed by surgery and surgery alone, four [47, 50, 53, 63]utilized a ​triple-combination neoadjuvant regimen​ (typically involving TACE, immune checkpoint inhibitors, and tyrosine kinase inhibitors), while one study [56] employed an ​immune checkpoint inhibitor (ICI)-based monotherapy. A ​sensitivity analysis​ was performed by excluding the study that used the ICI-based regimen. The subsequent meta-analysis of the remaining four studies—which specifically evaluated the triple-combination approach—revealed a ​significant improvement in OS and RFS​ among HCC patients receiving neoadjuvant therapy compared to surgery alone. These findings suggest that ​triple-combination neoadjuvant therapy​ may be particularly effective and could be considered a favorable strategy in the management of resectable HCC, especially in populations at high risk of recurrence.
Several limitations warrant attention in this study. Firstly, there was notable heterogeneity among the included studies in this meta-analysis. This heterogeneity could primarily be attributed to variations in patient baseline characteristics, varying definitions of resectability, and differences in conversion treatment regimens. To assess the robustness of the findings and provide further insights, sensitivity analyses were conducted. Secondly, the current landscape encompasses numerous types of ICIs and TKIs, yet the limited data available hindered a precise evaluation of the efficacy of each combination. Thirdly, the classification of unresectable HCC was not further subdivided, and the presence of macrovascular invasion and extrahepatic metastasis may influence the success rate of conversion and post-conversion therapy survival rates. Fourthly, although the primary outcome focused on the conversion rate, post-operative survival also held significant importance. Regrettably, only a few articles have reported the survival outcomes after surgical treatment in successfully converted patients, and these articles lacked sufficient follow-up duration. Fifthly, the limited articles and small samples in each article are also important limitations, which may cause bias. Hence, more studies with larger sample sizes are needed. Sixthly, most of the data in the conversion studies come from Asia, so whether the conclusions are applicable to Western populations remains to be further validated. Conclusion

Conclusion
Our study provides a comprehensive summary of the current evidence regarding the success rates of conversion therapy, including comparisons between different treatment regimens and associated adverse effects. Additionally, we present findings on the role and side effects of neoadjuvant therapy in resectable HCC.

Supplementary Information