A meta-analysis and real-world cohort study on the sex-related differences in efficacy and safety of immunotherapy for hepatocellular carcinoma

Background & Aims Sex-related differences in the immune pathogenesis of hepatocellular carcinoma (HCC), particularly related to oestrogen-dependent secretion of pro-tumourigenic cytokines, are well-known. Whether sex influences the efficacy and safety of immunotherapy is not known. Methods We performed a restricted maximum likelihood random effects meta-analysis of five phase III trials that evaluated immune checkpoint inhibitors (ICIs) in advanced HCC and reported overall survival (OS) hazard ratios (HRs) stratified by sex to evaluate sex-related differences in OS. In a real-world cohort of 840 patients with HCC from 22 centres included between 2018 and 2023, we directly compared the efficacy and safety of atezolizumab + bevacizumab (A+B) between sexes. Radiological response was reported according to RECIST v1.1. Uni- and multivariable Cox regression analyses were performed for OS and progression-free survival (PFS). Results In the meta-analysis, immunotherapy was associated with a significant OS benefit only in male (pooled HR 0.79; 95% CI 0.73–0.86) but not in female (pooled HR 0.85; 95% CI 0.70–1.03) patients with HCC. When directly comparing model estimates, no differences in the treatment effect between sexes were observed. Among 840 patients, 677 (81%) were male (mean age 66 ± 11 years), and 163 (19%) were female (mean age 67 ± 12 years). Type and severity of adverse events were similar between the two groups. OS and PFS were comparable between males and females upon uni- and multivariable analyses (aHR for OS and PFS: 0.79, 95% CI 0.59–1.04; 1.02, 95% CI 0.80–1.30, respectively). Objective response rates (24%/22%) and disease control rates (59%/59%) were also similar between sexes. Conclusion Female phase III trial participants experienced smaller OS benefit following ICI therapy for advanced HCC, while outcomes following A+B treatment were comparable between sexes in a large real-world database. Based on the ambiguous sex-related differences in survival observed here, further investigation of sex-specific clinical and biologic determinants of responsiveness and survival following ICIs are warranted. Impact and implications While immune checkpoint inhibitors have emerged as standard of care for the treatment of hepatocellular carcinoma, there are conflicting reports on whether the efficacy of cancer immunotherapy differs between females and males. Our study suggests ambiguous sex-related differences in outcomes from immunotherapy in hepatocellular carcinoma. Further investigation of sex-specific clustering in clinicopathologic and immunologic determinants of responsiveness to immune checkpoint inhibitor therapy should be prioritised. Systematic review registration PROSPERO CRD42023429625.


Introduction
Advanced-stage hepatocellular carcinoma (HCC) has historically been associated with a poor prognosis due to limited systemic treatment options.However, the treatment landscape has changed rapidly in recent years.After tyrosine kinase inhibitors (TKIs) dominated the field for over a decade, 1 the recent addition of immune checkpoint inhibitor (ICI)-based combinations of atezolizumab plus bevacizumab (A+B) and tremelimumab plus durvalumab (T+D) have increased options for first-line systemic therapy. 2,3While ICIs represent the mainstay of treatment across a wide variety of malignancies, there are conflicting reports as to whether the outcomes from cancer immunotherapy may differ between females and males. 4,5here are several immunological, biological, and behavioural differences between females and males that may affect efficacy and safety of immunotherapy.Sex-related differences in the regulation of innate and adaptive immune responses are known to play a role in hepatocarcinogenesis 6 and may also influence response to immunotherapy. 7According to preclinical studies, sex hormone-associated differences such as oestrogen-mediated inhibition of IL-6 expression reduced the risk of HCC development in female animals. 6][10] It has been postulated that male patients might derive a larger relative benefit from ICI than female patients since tumours in females may be less immunogenic and enriched with more potent mechanisms of immune escape than tumours in males. 4,11,12In addition, there are confounding behaviours (e.g., smoking) which may be unequally distributed between sexes displaying strong positive co-associations with increased tumour mutational burden and ICI efficacy. 4,13tudies exploring the interaction between patients' sex and the safety and efficacy of immunotherapy are scarce in patients with advanced HCC.
To fill this knowledge gap, we designed this study to systematically assess potential sex differences in overall survival (OS) in phase III clinical trials testing immunotherapy in advanced HCC.We also examined sex differences in the AB-Real study; a global, multicentre cohort of patients with HCC treated with A+B in routine clinical care.

Meta-analysis of phase III randomised-controlled trials
We fitted a restricted maximum likelihood random effects model including all available subgroup analyses of OS data in patients with HCC stratified by sex.Inclusion criteria were: (i) phase III randomised-controlled trials (RCTs) in the palliative treatment setting, (ii) evaluation of ICIs alone or in combination with other systemic agents, (iii) OS being a primary endpoint, and (iv) available subgroup analysis of OS stratified by sex.Studies evaluating loco-regional therapies as monotherapy or in combination with systemic treatments, as well as trials evaluating systemic treatments in a (neo)adjuvant setting, were excluded.
The literature search was restricted to studies published in English and conducted in MEDLINE (https://pubmed.ncbi.nlm.nih.gov), and Embase (www.embase.com) between 1 st of January 2007 and 21 st of May 2023.Conference abstracts published until scientific societies: the American Society of Clinical Oncology, the European Society of Medical Oncology, the European Association for the Study of the Liver, and the American Association for the Study of Liver Diseases.The complete search strategy is reported in the Supplementary Methods 1.The study protocol was registered in PROSPERO, an international prospective register of systematic reviews (registration code CRD42023429625; https://www.crd.york.ac.uk/prospero/#searchadvanced).
We screened 11,089 studies, leading to an identification of 10 phase III trials for analysis (Fig. S1).Sex-specific OS data was available in five clinical trials.We extracted hazard ratios (HRs) for patient sex subgroups from unstratified Cox proportional-hazards models with 95% CIs for OS.The meta-analysis was calculated using the 'metafor' package (https://cran.r-project.org/web/packages/metafor/metafor.pdf). 14A funnel-plot including all different studies (Fig. S2) shows a low probability of inclusion bias.
To investigate potential differences in treatment effect between male and female patients, we calculated the differences of the log HRs in male and female patients for each phase III study, as well as the corresponding standard errors.We then performed a random-effect meta-analysis to account for potential betweenstudy heterogeneity.A Forest plot of the random-effect metaanalysis is displayed in Fig. S3.
The AB-Real cohort of patients with HCC treated with atezolizumab plus bevacizumab Patients with histologically or radiologically diagnosed HCC who received A+B between May 2018 and January 2023 were included.Patients were retrospectively recruited by an international consortium including 22 centres from three different continents (Asia, Europe, and Northern America).Eligible patients were required to fulfil the following inclusion criteria: i) diagnosis of HCC by histopathological confirmation or imaging criteria according to the American Association for the Study of Liver Diseases 15 or the European Association for the Study of the Liver 16 guidelines, as well as ii) treatment initiation of A+B.Overall, the multicentre database included 840 eligible patients.Demographic and clinical data were collected retrospectively and curated at each participating centre.Ethical approval to conduct this study was granted by the Imperial College Institutional Review Board (Reference Number R16008).

Reporting of sex
As suggested by guidelines on reporting of sex, 17 the preferred terms used throughout this manuscript are sex, female and male sex.The sex of human research participants was defined based on self-reporting. 18

Study endpoints
This work aimed to determine differences in outcomes, treatment efficacy and safety aspects in female vs. male patients with HCC treated with A+B.Radiological response was evaluated by the treating physician according to RECIST v1.1 criteria.Disease control rate (DCR) was defined as the proportion of patients achieving stable disease (SD) or partial/complete response as best overall response (BOR), while objective response rate (ORR) reflected the proportion of patients with partial/complete response.The date of A+B initiation was considered as baseline for this study.Patients were followed until death or last follow-up (for censored patients) and patients alive at the data cut-off were censored at the date of the last clinical follow-up.
We aimed to document potential sex-related differences regarding baseline patient, tumour, and liver disease characteristics, and to evaluate efficacy (i.e., OS, progression-free survival [PFS], time to progression [TTP], BOR) as well as safety (i.e., adverse events [AEs]) according to sex.
Safety was reported as the incidence of AEs according to CTCAE version 4.0 or 5.0.The grading and causality of the AEs were assessed locally by the treating physicians.

Statistical analyses
Statistical analyses were performed using R 4.3.1 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria).All available patients fulfilling inclusion criteria were considered for this study.Data on baseline patient and tumour characteristics as well as radiographic features were summarised using descriptive statistics.Categorical variables were reported as absolute (n) and relative frequencies (%), while continuous variables were reported as mean ± SD or median (IQR), as appropriate.Student's t test was used for group comparisons of normally distributed variables and Mann-Whitney U test for non-normally distributed variables.Group comparisons of categorical variables were performed using either Chi-squared or Fisher's exact test, when the expected count in at least one cell was equal to or below 5.
OS was defined as the time from treatment initiation until death, and patients who were still alive or lost to follow-up were censored at the date of last contact.PFS was defined as time to radiological progression or death, whatever came first; patients alive or lost to follow-up without radiological progression were censored at the date of last contact.TTP was defined as time from treatment initiation until radiological tumour progression and only patients with available radiological re-staging were included in this analysis.Time on treatment was defined as the time from treatment start until end of treatment; patients who were alive or lost to follow-up with ongoing treatment were censored at the date of last contact.Median OS/PFS/TTP/time on treatment were calculated by the Kaplan-Meier method.Median estimated follow-up was calculated using the reverse Kaplan-Meier method. 19nivariable and multivariable analyses were conducted with Cox regression analyses.We also performed a subgroup analysis in those patients fulfilling the main inclusion criteria of the pivotal IMbrave150 phase III study (i.e., first-line treatment, Child-Turcotte-Pugh [CTP] stage A, Eastern Cooperative Oncology Group performance status [ECOG-PS] 0-1). 3he level of significance was set at a two-sided p value <0.05.

Results
Meta-analysis of phase III randomised-controlled trials Of 10 phase III RCTs identified, 2,20-28 only five studies reported OS according to sex and were therefore selected for this metaanalysis: IMbrave150, 20 HIMALAYA, 2 KEYNOTE-240, 22 LEAP-002, 21 RATIONALE-301. 26These tested the following treatments for advanced HCC, respectively: A+B vs. sorafenib, D+T or durvalumab monotherapy vs. sorafenib, pembrolizumab vs. placebo, pembrolizumab plus lenvatinib vs. lenvatinib plus placebo, and tislelizumab vs. sorafenib.Overall, 5,169 patients (n = 908 female and n = 4,261 male) were included in the analysis.Sorafenib was the control arm of all the studies, except for the KEYNOTE-240 trial, which tested pembrolizumab against placebo, and the LEAP-002 trial, which used lenvatinib in the control arm.The inclusion criteria appeared to be largely consistent between trials.
The OS benefit of the whole cohort (including females and males) was 20% (overall pooled HR 0.80, 95% CI 0.74-0.87;Fig. 1).A low degree of heterogeneity in the HR was indicated by I 2 = 0% and s 2 = <0.001 in the overall effect size model.In pooled subgroup analyses, findings revealed a significant survival advantage among male patients (pooled HR 0.79, 95% CI 0.73-0.86).While the pooled HR for female patients was only slightly higher than the one observed in males, the OS benefit did not reach statistical significance in females (pooled HR 0.85, 95% CI 0.70-1.03;Fig. 1).
The heterogeneity test of the differences between coefficients was in favour of homogeneity between the two groups and respective studies.The estimated treatment effect difference between sexes was 0.07 (95% CI 2.02−2.16),suggesting that no differences in the treatment effect between female and male patients were observed (Fig. S3; a positive value would correspond to a greater immunotherapy effect in males compared to females).

Differences in patient characteristics at study inclusion according to sex
Baseline characteristics, including mean age, BMI, prevalence of cirrhosis, liver function, BCLC stage, and performance status, were comparable between females and males, with only underlying aetiology being significantly different (Table 1).While the proportion of patients with viral hepatitis was higher in females (females: n = 82, 50%; males: n = 298, 44%), alcohol-related liver disease was more common in males (females: n = 12, 7%; males: n = 122, 18%; Table 1).

Discussion
Sex-related differences are known to affect innate and adaptive immune responses through a number of mechanisms including direct action of sex hormones on immune cell function, differential expression of genes located in sex chromosomes and altered epigenetic regulation of autosomal genetic material between the sexes.Sex-related diversity of nutritional status, gut microbial composition and disease-specific risk factors confer Fig. 3. Kaplan-Meier curves for overall survival, progression-free survival, and time to progression of patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab who fulfilled the main inclusion criteria of the IMbrave150 phase III trial according to sex.(A) Overall survival, (B) progression-free survival, and (C) time to progression of patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab who fulfilled the main inclusion criteria of the IMbrave150 phase III trial according to sex.
differential susceptibility to infection, autoimmunity, and response to vaccination. 7hether efficacy and safety of immunotherapy might be different between the sexes is the matter of contention.A metaanalysis of 20 RCTs of ICI treatment in advanced or metastatic cancers found improved OS with immunotherapy in male and female patients affected mainly by advanced melanoma and lung cancer, with enhanced magnitude of benefit shown in males. 4In contrast, an updated meta-analysis of 23 RCTs including 13,721 trial participants challenged this finding and reported similar OS across the sexes. 5None of the published studies included patients with advanced HCC, where sex-related differences have been shown to influence the pathogenesis and progression of cirrhosis and cancer. 6In our meta-analysis of five eligible phase III RCTs in advanced HCC, a statistically significant OS benefit for immunotherapy vs. control arm was demonstrated in the male subgroup (pooled HR 0.79, 95% CI 0.73-0.86),but not in female patients (pooled HR 0.85, 0.70-1.03).This is interesting, as female patients are commonly underrepresented in phase III trials evaluating novel systemic therapies in patients with HCC, but results are commonly extrapolated to both sexes and current guidelines do not differ between male and female patients. 1Importantly, when directly comparing model estimates of male and female patients, no differences in the treatment effect between sexes were observed.The RATIONALE-301 was the only trial where the mortality risk was significantly reduced by immunotherapy in female patients but not in males.The reasons for this observation are unclear, especially since this was not observed in the KEYNOTE-240 trial with pembrolizumab, which is also an anti-PD-1 antibody.Subgroups of phase III trials are not balanced for other prognostic factors; thus, it could well be that negative prognostic factors (e.g., macrovascular invasion, metastasis, high AFP etc.) were more common in the male subgroup.However, since baseline characteristics for males and females are not provided separately in the RATIONALE-301 or any of the other phase III trials included, this remains only speculative.
Next, we compared the efficacy and safety of A+B directly between male and female patients with HCC using AB-Real, the largest and most geographically heterogeneous study of patients treated with A+B in routine practice.The large number of female patients accrued to AB-Real (n = 163), which is 2-3 times higher than in ICI arms of phase III trials in advanced HCC, 2,20-28 allowed for robust analyses, complemented by thorough appraisal of prognostically relevant subgroups across sexes.In the AB-Real study, neither univariable nor multivariable models adjusted for other relevant prognostic factors revealed any differences in OS, PFS, and TTP between females and males.Ancillary measures of efficacy including ORR and DCR were also equal across groups.
0][31] Together, these findings do not suggest a significant association of patient sex with the efficacy of A+B in advanced HCC.However, it remains to be determined if this is specific to A+B or applies to other ICI mono-or combination therapies as well.Moreover, given the lack of a control group, we cannot appreciate whether the survival benefit derived from A+B might have been higher or lower than an alternative systemic therapy (i.e., TKI).
Autoimmune diseases are much more common in females, 32,33 thus one could speculate that female individuals undergoing immunotherapy are at higher risk of developing immune-related AEs.Indeed, in a recently published paper including an FDApooled analysis of landmark trials in HCC and a multiinstitutional dataset including over 357 patients with HCC treated with ICIs, the relative emergence of treatment-related AEs of grade > − 2 was higher in females. 34In contrast, we did not observe any differences in type, location, severity, or frequency of AEs in female compared to male patients.Notably, the FDA data analysis and multi-institutional cohort did not include patients treated with A+B, and the observed differences might have been driven by other treatments (e.g., anti-cytotoxic T-lymphocyte-associated protein 4). 34he meta-analysis has several limitations.Firstly, it was not based on individual participant data and subgroups were not stratified for other relevant prognostic factors, which could lead to imbalances between male and female subgroups and treatment arms.Secondly, several phase III RCTs could not be included as they did not report sub-analyses of OS stratified by sex.Thirdly, the subgroups are not balanced, with the sample size being larger in males, and the number of female patients in each trial being small.Finally, the phase III RCTs included were heterogeneous in terms of control arm and line of treatment.Therefore, the results of this meta-analysis can only be considered hypothesis-generating.
The real-world study has also some limitations beyond the well-known shortcomings of retrospective studies.The lack of a control group prevents conclusions on a potential higher relative benefit from immunotherapy vs. an alternative systemic therapy in females or males.The imaging schedule as well as modality was not pre-specified due to the real-world nature of the study.Since sex was based on self-reporting, it is possible that a very small proportion of included individuals might not be 46XX or 46XY; however, we assume that this would not have had a relevant impact on our results. 18n conclusion, while a slightly lower efficacy of immunotherapy in female patients with HCC was suggested in a metaanalysis of the sex-specific HRs for OS of five phase III RCTs, this could neither be confirmed when directly assessing the treatment effect differences between sexes in a similar metaanalysis, nor in a large global real-world cohort of patients treated with A+B, where efficacy and safety between males and females were similar.
Whilst treatment allocation based on patient sex is not recommended, our findings warrant continued investigation of sexrelated differences as a determinant of responsiveness to ICIs in patients with advanced HCC.

Fig. S2 .
Fig. S2.Funnel-plot displaying each study included in the meta-analysis according to sex, indicating low inclusion bias of included studies.Each dot represents a single study.The y-axis shows the standard error of the effect estimate, while the x-axis shows the observed outcomes for the respective studies.The outer dashed lines indicate the triangular region within which 95% of studies are expected to lie in the absence of both biases and heterogeneity.The dashed vertical line corresponds to no intervention effect Fig. S3

Fig. S3 .
Fig. S3.Forrest plot of a random effect meta-analysis of the differences between male and female coefficients of the Cox model of the phase III trials.Positive values correspond to a greater effect of immunotherapy in male compared to female patients Fig. S4

Table 2 .
Efficacy outcomes according to sex.
Categorical variables were reported as absolute (n) and relative frequencies (%).Group comparisons of categorical variables were performed using either Chi-squared or Fisher's exact test, as appropriate.* Compared by means of the log-rank test.
Categorical variables were reported as absolute (n) and relative frequencies (%), while continuous variables were reported as mean ± SD or median (IQR), as appropriate.Student's t test was used for group comparisons of normally distributed variables and Mann-Whitney U test for non-normally distributed variables.Group comparisons of categorical variables were performed using either Chi-squared or Fisher's exact test, as appropriate.P values in bold denote statistical significance.AFP, alpha-fetoprotein; ALBI, albumin-bilirubin; ArLD alcohol-related liver disease; BCLC, Barcelona Clinic Liver Cancer; CTP, Child-Turcotte-Pugh; ECOG-PS, Eastern Cooperative Oncology Group performance status; MASLD, metabolic dysfunction-associated steatotic liver disease.Fig.2.Kaplan-Meier curves for overall survival, progression-free survival, time to progression of patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab according to sex.(A) Overall survival, (B) progression-free survival, and (C) time to progression of patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab according to sex.

Table 3 .
Uni-and multivariable Cox regression analyses of factors associated with overall survival, progression-free survival, and time to progression.

Table 4 .
Comparison of treatment-emergent adverse events according to sex.

Table S1 .
Description of the trials included in the meta-analysis

Table S3 .
Uni-and multivariable Cox regression analyses of factors associated with overall survival, progression-free survival, and time to progression in patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab who fulfilled the main inclusion criteria of the IMbrave150 phase III trial.