Progression patterns and therapeutic sequencing following immune checkpoint inhibition for hepatocellular carcinoma: An international observational study

Abstract Background and Aims Different approaches are available after the progression of disease (PD) to immune checkpoint inhibitors (ICIs) for hepatocellular carcinoma (HCC), including the continuation of ICI, treatment switching to tyrosine kinase inhibitors (TKIs) and cessation of anticancer therapy. We sought to characterise the relationship between radiological patterns of progression and survival post‐ICI, also appraising treatment strategies. Methods We screened 604 HCC patients treated with ICIs, including only those who experienced PD by data cut‐off. We evaluated post‐progression survival (PPS) according to the treatment strategy at PD and verified its relationship with radiological patterns of progression: intrahepatic growth (IHG), new intrahepatic lesion (NIH), extrahepatic growth (EHG), new extrahepatic lesion (NEH) and new vascular invasion (nVI). Results Of 604 patients, 364 (60.3%) experienced PD during observation. Median PPS was 5.3 months (95% CI: 4.4–6.9; 271 events). At the data cut‐off, 165 patients (45%) received no post‐progression anticancer therapy; 64 patients (17.6%) continued ICI beyond PD. IHG (HR 1.64 [95% CI: 1.21–2.22]; p = .0013) and nVI (HR 2.15 [95% CI: 1.38–3.35]; p = .0007) were associated with shorter PPS. Multivariate models adjusted for progression patterns, treatment line and albumin‐bilirubin grade and Eastern Cooperative Oncology Group performance status at PD confirmed receipt of ICI beyond PD with (HR 0.17, 95% CI: 0.09–0.32; p < .0001) or without subsequent TKI (HR 0.39, 95% CI: 0.26–0.58; p < .0001) as predictors of prolonged PPS versus no anticancer therapy. Conclusions ICI‐TKI sequencing is a consolidated option in advanced HCC. nVI and IHG predict a poorer prognosis. Despite lack of recommendation, the continuation of ICI beyond progression in HCC is adopted clinically: future efforts should appraise which patients benefit from this approach.


| INTRODUC TI ON
While the dual programmed cell death ligand-1 (PD-L1)/vascular endothelial growth factor inhibitor combination has yielded undoubted superiority over sorafenib, not all patients are candidates for combination immunotherapy owing to concerns over toxicity.
Other immunotherapeutic approaches are available to sorafenibexperienced patients, such as pembrolizumab and nivolumab plus ipilimumab combination therapy (as well as previously nivolumab monotherapy), though survival benefit has been called into question, and access to these therapies is restricted to only certain healthcare  Conclusions: ICI-TKI sequencing is a consolidated option in advanced HCC. nVI and IHG predict a poorer prognosis. Despite lack of recommendation, the continuation of ICI beyond progression in HCC is adopted clinically: future efforts should appraise which patients benefit from this approach.

Key points
• Of the 364 studied patients, the median postprogression survival (PPS) after progressive disease on immune checkpoint inhibitors was 5.3 months.
• 45% of patients received no further anticancer therapy after disease progression.
• Continuation of immunotherapy after documented hepatocellular carcinoma (HCC) disease progression has been adopted in clinical practice, with 17.6% of patients in our cohort continuing treatment.
• In contrast to sorafenib, intra-hepatic HCC growth and new vascular invasion are associated with poorer outcomes after progression on immunotherapy.
• Both continuation of immunotherapy and treatment switching to tyrosine kinase inhibitors were associated with improved PPS. systems. 4 While the higher quality of evidence is available for molecularly targeted therapies following progression of disease (PD) on sorafenib, 5,6 there is no randomised trial evidence to suggest whether monotherapy or combination immunotherapy followed by second-line TKI leads to superior long-term survival outcomes compared to initial TKI use followed by either subsequent TKI or immunotherapy in the second line.
A key issue that complicates the optimal sequencing of systemic therapy is the heterogeneity of advanced HCC, where tumour progression and often contemporaneous progressive liver dysfunction dictate survival. As a result of this complex and dynamic interplay, there is high attrition from first-to second-line therapy in advanced disease: over 50% of patients who progress on first-line TKI are no longer eligible to further systemic therapy. 7,8 In patients who remain Different anticancer strategies are available to patients who experience PD on ICI, including treatment switching to TKIs. 9,10 Unlike sorafenib, a proportion of patients on ICI therapy can achieve significant disease downstaging, leading to the effective utilisation of locoregional or operative therapies in selected patients. 3,4 Additionally, it is becoming increasingly recognised that clinical benefit may be registered in patients who continue to receive immunotherapy beyond radiological PD, a practice that is consolidated in selected patients with non-small-cell lung cancer and melanoma; especially in the case of oligo-progressive disease. 11,12 Given the increasingly prevalent use of ICIs in the treatment of HCC, we designed this study with two aims. First, we intended to characterise the clinical features and radiological patterns of disease progression post-ICI and evaluate their relationship with clinical outcomes. Secondly, we aimed to retrospectively describe how different therapeutic sequencing strategies are associated with clinical outcomes in patients who, at the point of first radiological progression from ICI, remain eligible for further anticancer therapy.

| Study population
From an international consortium of 13 tertiary-care referral centres located in Europe, the United States and Asia, we accrued a prospectively maintained cohort of HCC patients undergoing treatment with ICIs between 2017 and 2021 (Table S1). Demographic and clinical data were collected retrospectively and curated at each participating centre. Eligible patients were required to fulfil the following inclusion criteria: (1) diagnosis of HCC by histopathological confirmation or imaging criteria according to the American Association for the study of Liver Disease 13 and European Association for the Study of the Liver 14 guidelines, (2) eligible for ICI monotherapy or combination therapy for HCC, not amenable to curative or loco-regional therapy  We then evaluated the relationship between PPS according to the treatment strategy at disease progression as follows: no post-progression anticancer treatment, ICI therapy beyond disease progression (ICI beyond PD) without subsequent TKI, postprogression TKI without continuation of ICI beyond PD, ICI beyond PD and subsequent TKI and other anticancer therapies (including systemic chemotherapy, LRTs and investigational agents within clinical trials). TKIs included sorafenib, regorafenib, lenvatinib and cabozantinib. As a receptor tyrosine kinase-targeting agent, ramucirumab was also included in this group. Post-progression continuation of ICI was defined as the continuation of treatment for more than 15 days after documented radiological PD. Switching to a different ICI (with or without a combination agent) at progression was not classed as the continuation of ICI beyond PD.

| Study end points
We reported the radiological patterns of disease progression and documented their distribution according to post-progression treatment strategies. Radiological patterns of progression were cat-

| Statistical analysis
Demographic data were summarised using descriptive statistics.
Median PPS was calculated using the Kaplan-Meier survival method and the median period of follow-up was calculated according to the reverse Kaplan-Meier method. Log-rank test was used for the univariable analysis of PPS. The chi-square test was used to compare categorical variables. Given the lack of censored data, the Mann-Whitney test was used to compare the time distribution for PFS of patients receiving and patients not receiving post-progression therapy. Cox proportional hazards regression was used for the univariable analysis of PPS according to the radiological patterns of disease progression, the multivariable analysis of PPS and to compute the hazard ratios (HRs) for death with 95% confidence intervals (CIs).
PPS curve plots were generated using the Kaplan-Meier method.
Cox regression survival probability plots for PPS were generated according to the presence of a given radiological pattern of progression. Each curve was obtained from separate multivariable models and superimposed, incorporating ECOG-PS at disease progression (0 and 1 vs. ≥2), ICI treatment line (1st vs. non-1st), ICI beyond PD and post-progression TKI exposure as adjusting factors.
The alpha level for all analyses was set to p < .05. All statistical analyses were performed using the MedCalc Statistical Software version 19.3.1 (MedCalc Software Ltd, Ostend, Belgium; https:// www.medca lc.org; 2020).

| Patient characteristics at immune checkpoint inhibitor initiation and at disease progression postimmune checkpoint inhibitor
At the data cut-off, 604 patients were entered into the study, with a    Table 3). Figure S2 describes PPS estimates for patients receiving ICI-based regimes as first-line therapy, categorised into the same groups.  Table S2.

| New vascular invasion and intra-hepatic tumour growth predict poorer post-progression survival at disease progression
Clinical and radiological features at disease progression may influence clinician behaviour as well as survival outcome. We therefore verified the relationship between these features and treatment strategy at PD on ICI: ECOG-PS and ALBI grade at PD were found to be significant determinants of treatment strategy (Table 2).

| DISCUSS ION
The systemic therapy armamentarium of HCC now recognises multiple therapeutic options across lines of therapy. Widening therapeutic options pose significant challenges to clinicians treating HCC patients, especially in view of the lack of biomarkers that can facilitate the selection of the most appropriate therapy a priori. 10,16 In this large, multi-institutional observational study of 364 patients, largely Our study portrays a substantial difference in the prognostic value of post-progression radiological features compared to the sorafenib era, where NEH and nVI led to the subclassification of the BCLC stage based on characteristics evident upon progression on sorafenib. 21 It should be acknowledged that NEH and nVI were categorised together in previous analyses of post-progression outcomes following treatment with sorafenib, 7 as well as ramucirumab 22 : we chose to evaluate these two patterns of progression separately, postulating that they could reflect different underlying tumour biology.
It is plausible that differences in the mechanism of action between TKIs and immunotherapy might explain the difference in the prognostic weight of these variables. Immunotherapy is in fact more likely than sorafenib to produce radiologically appreciable disease responses and is capable of leading to long-term disease stabilisation in up to 20% of patients with HCC. 3,4 Although patients who respond to immunotherapy tend to be those that benefit most in terms of survival, 23 evolving experience in the use of PD-(L)1 inhibitors in indications other than HCC suggests the potential for treatmentinduced benefit even in those patients who fail to achieve a significant radiological response; a finding that does not apply to TKIs.
When evaluating the clinical use of ICI in routine clinical practice, we were able to describe that >50% of patients received subsequent anticancer treatment after progression on ICI-based therapy: a higher percentage compared to the 23.6% of patients receiving second-line therapy in previous studies. 24 In an increasingly complex treatment landscape, our study shows that sequential exposure to multiple agents characterised by diseasemodulating activity in HCC is associated with incremental survival benefits in patients who remain fit for treatment, a consolidated concept from the TKI era in HCC. 25 18). ALBI grade at PD was however a determinant of specific post-progression strategy ( Table 2, p = .046). Why liver function at the commencement of ICI rather than at progression is associated with the overall receipt of further therapy in this study is not clear.
Prior therapy is a possible confounding factor herein: sorafenib pre-treatment was more common in patients not receiving postprogression therapy (57.0% vs. 49.7%). However, this difference was non-significant (p = .17, Table 1) and treatment line of ICI itself was not associated with receipt of post-progression therapy (p = .35), or therapeutic strategy at PD (p = .52, Table 2).
Regarding other potential determinants of receipt of postprogression therapy, it could be argued that adverse disease biology or intrinsic treatment resistance may preclude eligibility, or influence the clinician decision to treat further. However, PFS from ICI initiation was not significantly lower in patients who did not receive further lines of therapy (median 2.8 months vs. 3.5 months, p = .11, Table 1). Patients not receiving post-progression therapy were more commonly treated with ICI in combination with a second agent than those who did receive further therapy (p = .0005). This could reflect reduced availability of other drug classes or greater treatment toxicity. Prior liver resection was more common in patients receiving post-progression therapy than those who did not (37.7% vs. 24.2%, p = .0061). This may reflect a low burden of relapsed disease detected on surveillance following curative-intent surgery.
In our study, 17.6% of patients who progressed on ICI continued immunotherapy treatment beyond initial evidence of radiological progression, mostly in the context of the growth of existing lesions.
The retrospective, non-randomised nature of our data cannot constitute a platform for practice-changing recommendations. However, the median PPS of 15.3 months achieved by patients whose change to TKI therapy was deferred beyond initial evidence of progression to ICI is a provocative finding and compares favourably to survival estimates of patients who were treated with immediate switching to TKI. Continuation of ICI beyond the first point of progression is commonplace in clinical practice and often permitted in clinical trials of ICI, recognising the potential for cancer immunotherapy to alter tumour biology beyond strict radiological criteria 29 and allowing, in other tumour types, deferred switch to potentially less tolerable strategies such as chemotherapy 30,31 or TKIs. 32 To our knowledge, our study is the first to report on the clinical use of immunotherapy beyond RECIST progression in HCC and suggests that this strategy should be investigated in a subset of patients with advanced disease, potentially based on the radiographic characteristics of progression. The dissociation between radiological response and PPS in patients treated with ICI highlights the need for further research into modified imaging criteria and alternative immune monitoring biomarkers to identify patients who may derive greater benefit from immunotherapy. 33 Limitations to this study should also be acknowledged. The retrospective nature of the dataset, although prospectively maintained, limits the ability to draw definitive conclusions about post-progression outcomes, and the benefit of this approach over immediate switching to other effective therapies remains to be evaluated prospectively. Despite recruitment to the study being multicentre, selection bias needs to be considered and clinical practice at non-participating sites may be different. Prescribing practice (and consequently patient outcomes) may be skewed by differences in treatment availability and reimbursement policies across countries.
Pembrolizumab is approved as a second-line agent in the USA but not in the UK, which should be acknowledged as a potential confounding factor. Tumour imaging was also reviewed locally rather than centrally. Efforts were made to reduce selection bias through the employment of multivariable analyses, but the potential remains for other variables to confound associations with PPS.
Despite these potential limitations, this study gives a contemporary portrait of how immunotherapy has integrated the management of advanced HCC. In spite of the heterogeneity in prognosis, partly dictated by the diverse radiological patterns of disease progression, more than 50% of patients with HCC who experience PD in ICI receive subsequent therapy. Prospective research efforts should validate the impact of radiological patterns of disease progression and continuation of ICI post-radiological PD in patients with advanced HCC in order to optimise treatment sequencing in immunotherapyresistant disease.

ACK N O WLE D G E M ENTS
The authors would like to acknowledge the infrastructural support provided by Imperial Experimental Cancer Medicine Centre, Cancer Research UK Imperial Centre and the NIHR Imperial College BRC.