Impact of Neoadjuvant Immunotherapy on Recurrence-Free Survival in Patients with High-Risk Localized HCC

Hepatocellular carcinoma (HCC) accounts for 80% of all primary liver cancers and is currently the fifth leading cause of cancer mortality in the United States (1), with mortality closely matching incidence (2). Although the burden of HCC is variable by country, recent projections suggest that the worldwide incidence of liver cancer will continue to rise (3, 4), with the epidemiology of underlying liver disease shifting away from viral hepatitis and toward metabolic dysfunction-associated steatohepatitis (5). Only approximately 30% of patients diagnosed with HCC are considered eligible for resection by current Western guidelines due to the presence of extrahepatic and vascular extensions, poor hepatic reserve, and/or anatomical considerations (including multinodular disease) that preclude adequate resection margins (6–8). Even in those who undergo curative-intent resection, long-term outcomes remain poor, with a vast majority of patients recurring within 5 years of surgery (9–11), stemming from the presence of micrometastatic disease at the time of surgery and/or new foci of disease developing in the cirrhotic liver. Perioperative systemic therapy for HCC is an area of active clinical investigation with the potential to reduce micrometastatic disease burden and in turn improve postresection outcomes.

Multiple novel immunotherapy-based regimens have demonstrated improved survival versus sorafenib in unresectable HCC and expanded the vista of therapeutic options (12, 13). The impact of immune checkpoint inhibitor (ICI) therapy on the treatment landscape of advanced HCC has led to an intense interest in investigating its potential role in earlier stages of HCC. Recently, immunotherapy-based combinations have been shown to improve recurrence-free survival (RFS) in high-risk patients following curative-intent resection or ablation, although longer follow-up is needed to assess its impact on overall survival (OS; refs. 14, 15). In addition, multiple single-institution clinical trials have explored the feasibility of neoadjuvant immunotherapy (16, 17). Collectively, available studies of neoadjuvant immunotherapy in HCC show that this approach is feasible, including in those with high-risk tumors who were poor candidates for upfront resection, with pathologic response rates of 20% to 33% (18–20). Whether neoadjuvant immunotherapy can improve long-term survival outcomes is unknown. We hypothesize that patients who undergo neoadjuvant ICI treatment have higher-risk tumor features compared with those who undergo upfront surgery but that neoadjuvant ICI may allow for such high-risk patients to achieve comparable survival outcomes.

We retrospectively identified 95 patients who underwent liver resection at the Johns Hopkins Hospital between January 1, 2017, and December 1, 2023, with a pathologically confirmed diagnosis of HCC. Electronic medical records were reviewed to determine demographic characteristics, clinical features, laboratory values prior to initiation of neoadjuvant ICI–based therapy or surgery (whichever occurred first), treatment history (in particular, receipt of neoadjuvant ICI therapy), tumor characteristics based on pathology of resected tumor specimen, and clinical outcomes. The primary clinical endpoints of interest included RFS (defined as time from curative-intent hepatectomy to radiographic disease recurrence or death due to any cause) and OS (defined as time from curative-intent hepatectomy to death due to any cause). Data cut-off was set as January 3, 2024. This study was approved by the Johns Hopkins Institutional Review Board (#00417376) and performed in accordance with the U.S. Common Rule.

Our institutional multidisciplinary treatment practice patterns evaluate patients for resection who are outside of traditional resection criteria established by the Barcelona Clinic Liver Cancer (BCLC) Staging System (21), but usually only after a period of systemic therapy. Since the advent of effective combination systemic therapies, our institution has offered resection to patients with high-risk tumor features following neoadjuvant therapy who previously were not considered for upfront resection. Since 2017, our institution has prospectively enrolled patients on neoadjuvant clinical trials of anti-PD1–based systemic therapies [including cabozantinib plus nivolumab (NCT03299946) and nivolumab alone or in combination with lymphocyte activation gene-3 (LAG3) inhibitor relatlimab (NCT04658147)]. These clinical trials specifically enabled patients outside of traditional Western resection criteria to enroll, including patients with multifocal disease, locally advanced disease, and portal vein invasion. Whereas resectability in localized HCC is defined not only by the likelihood of achieving a margin-negative (R0) resection but also by the safety of the surgery in the cirrhotic liver and the extent of high-risk disease features (e.g., vascular invasion), upfront resectability is defined in this analysis according to the BCLC criteria of BCLC A disease with a solitary lesion (22).

Patients treated according to our various study protocols, as well as patients treated with systemic therapy off-protocol, were both included in the primary analysis. Mixed phenotypic tumors (e.g., mixed HCC/cholangiocarcinoma) were excluded from the present analysis. Patients who did not undergo curative-intent hepatectomy or had metastatic disease identified during surgery that was previously unrecognized, were excluded. Patients who had combination locoregional therapy and local therapies (e.g., transarterial chemoembolization/transarterial radioembolization; Y90) were included in the overall cohort but not included in the analysis for the effect of pathologic response on RFS in the neoadjuvant immunotherapy treatment arm if they received the local therapy following immunotherapy because of an inability to attribute pathologic responses to immunotherapy alone. Inclusion and exclusion criteria are further outlined in Supplementary Fig. S1.

We applied univariate Cox proportional-hazards survival models to assess effect sizes and generate HR based on demographic and tumor-specific features. Kaplan–Meier survival curves were constructed to compare survival between various clinical cohorts, including neoadjuvant ICI–treated and untreated patients, with significance determined by the two-sided log-rank test. Differences between categorical variables were assessed using the Fisher exact test, whereas those between numeric variables were assessed using the Mann–Whitney U test.

The data generated in this study are available upon request from the corresponding author.

The final clinical cohort included a total of 92 patients, 36 of whom underwent neoadjuvant ICI–based treatment (Table 1). The overall cohort was predominantly male (69.6%), White (57.6%), with preserved liver function (98.9% Child Pugh A). Of those who received neoadjuvant ICI (Table 2), a majority were treated with anti-PD1–based therapy, either as monotherapy (27.8%), in combination with tyrosine kinase inhibitor (36.1%), or in combination with anti-LAG3 (16.7%). A majority of those treated with neoadjuvant ICI were treated under a clinical trial protocol (69.4%) for at least two cycles. Those who received neoadjuvant ICI–based treatment were more likely to have higher-risk disease, as indicated by the greater proportion of patients who had α-fetoprotein (AFP) ≥ 400 ng/mL at baseline (38.9% vs. 14.3%, Fisher exact test P = 0.02), large tumors greater than 5 cm (72.2% vs. 37.5%, Fisher exact test P = 0.001), portal vein invasion (25.0% vs. 0%, Fisher exact test P < 0.001), and multiple tumor foci (50.0% vs. 12.5% multiple tumors, Fisher exact test P < 0.001). Most patients (83.3%) treated with neoadjuvant ICI had stable disease by RECIST version 1.1 as the best response to therapy prior to surgery, with a minority (13.9%) achieving partial response. In both cohorts, a comparable minority of patients received local therapy prior to surgery. Twelve (33.3%) patients in the neoadjuvant ICI–treated cohort received adjuvant therapy, most commonly in the form of anti-PDL1 monotherapy.

We first sought to determine the impact of neoadjuvant ICI on long-term outcomes. The median RFS in the ICI-treated cohort was 44.8 months [95% confidence interval (CI), 38.4–not reached, (NR)] compared with 49.3 months in the upfront surgery cohort (95% CI, 27.8–NR); log-rank P = 0.66 (Fig. 1A). The median OS has not been reached in both cohorts; log-rank P = 0.97 (Fig. 1B). We also evaluated the patterns of recurrence for patients who recurred in both cohorts (Fig. 2) and observe that neoadjuvant ICI–treated patients more commonly had distant recurrences.

In patients treated with neoadjuvant ICI who did not undergo liver-directed local therapy following ICI treatment, 10/33 (30.3%) patients had a major pathologic response at the time of resection, defined as tumor necrosis ≥70%. In contrast, 6/33 (18.2%) had a minor pathologic response (necrosis 30%–69%) and 17/33 (51.5%) had no pathologic response (necrosis 0%–29%). Patients who achieved a major pathologic response at the time of surgery had numerically superior RFS compared with those with a minor response (median NR vs. 38.3 months, log-rank P = 0.16, Fig. 3A). In a multivariable univariate Cox regression, a major pathologic response was not predictive of RFS (HR 0.62, 95% CI, 0.03–11.03) when controlled for variables, including albumin–bilirubin (ALBI) grade, AFP ≥ 400 ng/mL, tumor size ≥ 5 cm, tumor focality, R0 resection status, vascular invasion on final pathology, and tumor grade. Upfront resectability status was negatively prognostic for RFS (HR 4.39, 95% CI, 0.82–23.53) but did not reach statistical significance. ALBI grades 2 and 3, tumor size ≥ 5 cm, and the presence of vascular invasion on final pathology were negative predictors for RFS in the ICI-treated cohort (Fig. 3B).

In the total cohort of neoadjuvant ICI–treated patients, 22/36 (61.1%) patients were outside of traditional resectability criteria due to multifocal tumors or vascular invasion. In particular, nine (25.0%) patients had BCLC C disease with evidence of portal vein invasion. Nevertheless, a vast majority of patients (94.4%) were able to undergo successful R0 resection, comparable with that of patients who underwent upfront surgery. Detailed pathologic outcomes in both subgroups are outlined in Table 3. Figure 4 illustrates CT images three of such patients at baseline, following completion of neoadjuvant ICI therapy, and following successful resection.

We report long-term clinical outcomes of patients treated with neoadjuvant ICI–based therapy and a contemporaneous cohort of those who underwent upfront resection for HCC at a single academic institution. The patients who received neoadjuvant ICI therapy were generally patients who had tumors that were deemed high-risk for upfront resection due to large tumor size, proximity to critical structures, multifocality, or suspected macrovascular invasion. Conversely, the majority of patients in the upfront surgery cohort had solitary lesions and would generally have received upfront resection under current BCLC guidelines. We observe that that whereas most patients who underwent neoadjuvant therapy fell outside of traditional BCLC criteria for resection, the RFS in those who received neoadjuvant ICI therapy was comparable with those who underwent upfront surgery.

These data reflect the clinical practices of the Johns Hopkins Liver Multidisciplinary Clinic, where patients with HCC outside of conventional resection criteria are often considered for upfront systemic therapy or locoregional therapies (including transarterial chemoembolization/transarterial radioembolization, Y90, and radiation therapy). Although many factors are considered, in general, patients who are deemed to be unresectable upfront with macrovascular invasion and/or larger tumor burdens, as reflected by extremely high AFPs (i.e., >10,000 ng/mL) or large tumors (i.e., >10 cm), are more likely to be offered systemic therapy. Notably, although prior studies have reflected on the feasibility of resection on patients with features such as large tumor size, multinodular disease, and major vascular invasion, outcomes were inferior compared with those falling under traditional resectability criteria (23). Our observation that RFS was comparable in the ICI-treated and untreated cohorts provides initial evidence that neoadjuvant ICI–based therapy may be effective in transforming the natural histories of these patients postresection to that of one comparable with those who received upfront surgery for lower-risk HCC. However, larger prospective studies are needed to confirm that neoadjuvant immunotherapeutic strategies can expand the pool of patients who should be considered for resection.

Several neoadjuvant and perioperative ICI-based studies have demonstrated that single and combination ICI therapy is well tolerated, with a vast majority of patients undergoing successful resection and some observations suggesting that the presence of a major pathologic response at the time of surgery may predict recurrence (18–20). However, given that the study of neoadjuvant ICI–based therapy in HCC is still in its relative infancy, the optimal endpoint for neoadjuvant investigations that translates to a change in the natural disease trajectory of resected HCC remains unclear. Our results suggest that upfront resectability status was a negative prognostic factor for RFS, although not statistically significant. Furthermore, patients with a major pathologic response (i.e., ≥70% necrosis) at the time of surgery had numerically superior RFS. Although there seems to be a correlation, the threshold of tumor necrosis as it translates to improved survival outcomes remains under active investigation. In addition, the role of adjuvant ICI remains relevant given the observation that more distant recurrences occurred in the neoadjuvant ICI–treated cohort, underscoring the need for systemic control.

Due to the retrospective nature of this analysis, there are several limitations to consider with the interpretation of these results. We cannot exclude the possibility that differences between the lower-risk upfront surgery and higher-risk neoadjuvant cohorts are the reason why no difference was observed in their clinical outcomes. For example, we note that although the collected demographic characteristics of the cohorts were largely comparable, a numerically higher proportion of Black patients were in the upfront surgery cohort. This difference could be related to patient-level factors or comorbidities affecting neoadjuvant clinical trial eligibility, which in turn affected HCC outcomes. The true effect size for the impact of neoadjuvant ICI therapy is difficult to estimate without a similarly high-risk group of patients who underwent upfront surgery, which is typically not undertaken in the standard-of-care management of HCC. In addition, patients who underwent neoadjuvant ICI–based therapy as described in this article does not represent the experience of all patients who were treated with neoadjuvant ICI regimens because only those who then underwent resection were included. A small minority of patients treated in the neoadjuvant trials represented in this analysis were ultimately unable to undergo surgery, most commonly due to the presence of more extensive disease than what was recognized. Finally, our single-institution experience may not be applicable to other centers due to differences in institutional practice and patient populations. Despite these limitations, this is the largest retrospective cohort to report on the outcomes of neoadjuvant ICI–treated and untreated patients in a rapidly evolving landscape of ICI-based treatments available in HCC.

Although patients treated with neoadjuvant ICI–based therapy for HCC in this cohort had higher-risk disease features and generally would not have been considered surgical candidates under BCLC criteria, they achieved long-term outcomes comparable with a cohort of patients who underwent upfront surgery. Our observations highlight the need for future prospective trials to further defined the role of neoadjuvant ICI therapy in both traditionally resectable and high-risk localized HCC populations.

M. Nakazawa reports grants from the NIH during the conduct of the study and grants from Lou and Nancy Grasmick Fellowship, Linda Rubin Pancreatic Cancer Fellowship, and James and Frances McGlothlin Fellows to Faculty Award outside the submitted work. R.A. Anders reports grants and personal fees from Bristol Myers Squibb, personal fees from AstraZeneca, personal fees from Merck, grants from RAPT Therapeutics, grants from Break Through Cancer, and personal fees from JAZZ Oncology during the conduct of the study. A.K. Kim reports personal fees from AstraZeneca outside the submitted work. J. Meyer reports other support from Boston Scientific, personal fees from Springer, and personal fees from UpToDate outside the submitted work. K. Hong reports grants and personal fees from Boston Scientific, grants from Merit Medical, and personal fees from Varian outside the submitted work. M. Baretti reports personal fees from AstraZeneca and Incyte outside the submitted work. A.T. Strauss reports grants from the NIH National Institute of Diabetes and Digestive Kidney Diseases during the conduct of the study. M. Yarchoan reports grants and personal fees from Genentech, grants from Bristol Myers Squibb, grants and personal fees from Exelixis, grants from Incyte, grants from Eisai, personal fees from AstraZeneca, personal fees from Replimune, personal fees from Hepion, personal fees from Lantheus, and other support from Adventris outside the submitted work. No disclosures were reported by the other authors.

We thank the patients and their families and all the providers and site personnel who contributed to the care of these patients. M. Nakazawa is supported by the NIH T32 CA009071-42 training grant, the Lou and Nancy Grasmick Fellowship, the Linda Rubin Pancreatic Cancer Fellowship, and the James and Frances McGlothlin Fellows to Faculty Award. A.T. Strauss is supported by grant number K08DK133638 from the National Institute of Diabetes and Digestive and Kidney Diseases. This study was supported by the NCIs Specialized Program of Research Excellence in Gastrointestinal Cancers (P50 CA062924), the NIH Center Core Grant (P30 CA006973) and R01CA285544 (to W.J. Ho and M. Yarchoan), and the Johns Hopkins Bloomberg–Kimmel Institute for Cancer Immunotherapy. The authors wish to acknowledge the Clinical and Translational Research Unit of the Division of Gastroenterology and Hepatology at Johns Hopkins University for the use of the Liver Disease Precision Medicine Analytics Platform.