Finn Sheds Light on Single-Agent Pembrolizumab Data in Advanced HCC

Richard S. Finn, MD, discusses results from the KEYNOTE-240 trial and explained how they are still clinically meaningful for patients with hepatocellular carcinoma.

Richard S. Finn, MD

Although it did not meet its primary endpoints, pembrolizumab (Keytruda) was found to show a benefit in survival for patients with advanced hepatocellular carcinoma (HCC) who were previously treated with sorafenib (Nexavar) in the phase III KEYNOTE-240 trial—and the results are still clinically meaningful, explained Richard S. Finn, MD.

Data from a final analysis of the placebo-controlled KEYNOTE-240 study, which evaluated the PD-1 inhibitor in patients with advanced HCC who received prior sorafenib, were presented at the 2019 ASCO Annual Meeting. KEYNOTE-240 was the confirmatory trial for the FDA's accelerated approval in November 2018 for pembrolizumab in this setting, which was based on the phase II KEYNOTE-224 trial.

In KEYNOTE-240, 413 patients were randomized 2:1 to receive either pembrolizumab plus best supportive care or placebo and best supportive care. At a median follow-up of 13.8 months, both progression-free survival (PFS) and overall survival (OS), the co-primary endpoints of this study, were improved.1 However, these findings did not meet statistical significance per the prespecified protocol criteria. There was a 22% reduction in the risk of death with pembrolizumab (HR, 0.78; 95% CI, 0.611-0.998; P = .0238), while the objective response rate (ORR) remained consistent with prior data from the KEYNOTE-224 study.

KEYNOTE-224 was a phase II trial that investigated single-agent pembrolizumab in patients with advanced HCC who either progressed on or became intolerant to sorafenib.2 Of the 104 patients included on this trial, the ORR was 17%. The median PFS was 4.9 months, and the median OS was 12.9 months.

Grade 3 treatment-related adverse events (TRAEs) were noted in 24% of patients, while the most common TRAEs were increased aspartate aminotransferase, increased alanine aminotransferase, and fatigue. Investigators from the KEYNOTE-240 trial noted that the safety profile was similar to that of the KEYNOTE-224 trial and of other pembrolizumab trials.

In an interview with OncLive, Richard S. Finn, MD, professor of clinical medicine, Department of Medicine, Division of Hematology/Oncology, Geffen School of Medicine, University of California, Los Angeles, and Director of the Signal Transduction and Therapeutics Program at the Jonsson Comprehensive Cancer Center at UCLA, discussed the clinically meaningful findings from the KEYNOTE-240 trial.

OncLive: What was the rationale for investigating pembrolizumab monotherapy in the second-line setting for HCC?

Finn: KEYNOTE-240 was a phase III randomized, placebo-controlled study of the PD-1 antibody pembrolizumab versus best supportive care in second-line HCC. Obviously, PD-1 inhibitors have been used in many diseases and have had a big impact, so the rationale here comes from 2 things. First, 90% of the time liver cancer arises in the context of some background liver disease andan inflammatory state. There is some immune-intolerance that develops as a result of chronic inflammation, and that can contribute to the development of liver cancer. In addition, there were 2 phase II studies that demonstrated fairly good responses in liver cancer with PD-1 directed therapies

First, the CheckMate 040 study with nivolumab was a phase II study for both first- and second-line patients,The second-line patients had a response rate of roughly 15%. The patients who responded had a very long duration of response; median duration of response was over 16 months. These data are not something you typically see in liver cancer with the tyrosine kinase inhibitors (TKIs), which are approved in the first and second line therapy. However, when this study was launched, there was nothing approved for the second-line in HCC.

The KEYNOTE-224 was another single-arm phase II study of pembrolizumab in a similar population of second-line patients. There again, we found a response rate of around 15% and the median duration of response that was very long, well over 1 year. Both nivolumab and pembrolizumab have accelerated approvals for second-line in the United States. The KEYNOTE-240 study was a phase III study [designed] to prove, in a randomized setting, that these drugs can improve outcomes for these patients.

How was the KEYNOTE-240 trial designed?

KEYNOTE-240, like I said, was a randomized, placebo-controlled, phase III study. Patients were required to have had prior sorafenib, whether they progressed or became intolerant to it. As it turns out, an overwhelming majority of patients were progressors on sorafenib. They had to have Child-Pugh A liver function, and we excluded patients who had invasion of the main portal vein also known as the extrahepatic portal vein as these patients then to have a very poor prognosis.

Patients were randomized 2:1 to pembrolizumab versus placebo, and the study had dual endpoints. One was PFS at the first interim analysis, and the other endpoint was OS. What is important to understand is that the statistical design of the study required a pretty high stringency to be called positive. What I mean by that is the alpha, the P value for significance for the amount of Type 1 error (false-positives) that would be accepted, was split initially between PFS and OS, because those were the dual primary endpoints. We also had 2 interim analyses for OS before we did the final analysis, which was presented at ASCO. These additional analyses also required us to lower the P-value to call statistical significance.

Every time we looked at the interim data, the study team would have to subtract some of the alpha from what you had started with. At the end of the day, the final analysis the P value for OS had to be 0.017, and for PFS it had to be 0.001, so [it was a] very high statistical bar to be called positive. In most studies, the P value is 0.05. If you’re less than 0.05, you’re positive, but because of the way this study was designed, you were penalized a little bit, so the P value goes down.

What were the findings presented for the final analysis?

We presented [at ASCO] the final results from the study. As far as PFS, the median went from 3.8 months to 4.2 months with a hazard ratio (HR) of 0.71, the confidence interval (CI) was less than 1, but the P value was 0.02, which was higher than the 0.001 that was required to be positive. The PFS was looked at in the first interim analysis, and what you see over time is you is that the PFS curve remains fairly flat, where we have about 16% of patients who never progressed as far as we followed them. This is typical for a lot of the immune checkpoint studies in other diseases where the median PFS doesn’t necessarily capture the whole treatment effect.

For OS, it was impressive. It was 10.6 months in the control arm and 13.9 months in the treatment arm. That’s more than a 3-month improvement for the median OS. The upper limit of the CI was 0.998, and the HR was 0.78. [There was] a 22% decrease in the risk of death. Again, the P value here was 0.0238, but to be positive, it had to be 0.017. Still, this is a very strong result for OS, but because of the required P value, we can’t say it was a positive study since it did not meet its primary endpoint.

This is a very clinically meaningful result, especially when we look at the response rate, which was 18.3%. The median duration of response was 13.8 months. That’s what we hope for in the immuno-oncology space. It would be great if it were like in melanoma where the response rate is around is well over 50%but this really confirmed the results of the phase II studies in many ways. These are the results that led to the accelerated approval in terms of duration of response.

For toxicity, there’s really nothing new here. The adverse effect profile for pembrolizumab in this population was very similar to what’s been seen in other diseases. We also did an analysis of the impact of subsequent therapies and how that could impact the OS results as there were no approved second line agents when the study was launched but during the course of the study regorafenib was approved, we had other phase III data, as well as the approval of immunotherapies by accelerated approvals, so we did sensitivity analyses for patients who went on to get other treatments. In those analyses, we had very consistent results of a HR of about 0.65 which was closer to our initial assumptions. This suggests that part of the reason we didn’t meet our statistics is because patients went on to other treatments.

Do you have any next steps planned for the KEYNOTE-240 data?

Currently pembrolizumab has an accelerated approval, and KEYNOTE 240 confirms the data that got it approved though the study, strictly speaking was negative. I don’t think they will pull the drug because it’s an activedrug. There is another phase III study that has a very similar design in Asia that has not read out yet and the results of the study should also be confirmatory. It will be very interesting to see how this compares to where they don’t have access to many other drugs after progression, especially checkpoint inhibitors, and to see if that study is positive. It could be very supportive, and I think if you have a phase III study that is clearly positive and then KEYNOTE-240, that makes a strong case to try and get a full approval.

Meanwhile, there are other drugs with pembrolizumab. There’s a front-line study of pembrolizumab plus lenvatinib (Lenvima) in combination; that’s a phase III study. It’s based on the combination which has shown a response rate instead of being 18% with pembrolizumab alone, we’re seeing response rates around 30% or higher with the combination. That has been moved to the frontline study versus lenvatinib.

What do you think is important to take away from these data?

Finn: Clinically, I think the results are very meaningful. The odds of this study being a false-positive result, which is the whole point of the alpha level; when we say there is an alpha of 0.05, we’re saying we’re willing to accept a 5% false-positive. The odds of the results of KEYNOTE-240 occurring by accident is very low. Statistically, this is actually a very likely result that is clinically very meaningful. Sometimes you see a P value around 0.07, but that CI goes over 1. Here, the CI is less than 1 so it’s a fairly tight result.

We did present at the ESMO World Congress on Gastrointestinal Cancer an updated analysis looking at survival by region. In all the major regulatory regions, such as the European Union, North America, and Asia, the HRs were all less than 0.77. Then there were some other countries that were not part of these region that had a HR well above 1, like 1.89, which is an interesting observation. When you do a global study, certain practice patterns might affect outcomes.

Is there anything else important to note in the treatment landscape for HCC?

Finn: There’s a lot of important stuff. There’s a lot of novel studies going on. Now that we have established the activity of single agent nivolumab and pembrolizumab, with response rates of 15% to 20% the challenge is how do we improve this. We don’t have any biomarkers to identify those patients. The strategy is to increase the number of patients who benefit. This is why there is interest in combining immunotherapies with the TKIs, whether it’s cabozantinib or regorafenib or lenvatinib. Also, there is interest in combining checkpoint inhibitors, PD-1 with CTLA-4, and also combining PD-1 inhibitors with anti-VEGF approaches. There’s a lot of activity going on, and it’s a very serious disease. There’s a lot to think positively about and we have to keep trying.


  1. Finn RS, Ryoo BY, Merle P, et al. Results of KEYNOTE-240: phase 3 study of pembrolizumab (Pembro) vs best supportive care (BSC) for second line therapy in advanced hepatocellular carcinoma (HCC). J Clin Oncol. 2019;37(suppl; abstr 4004).
  2. Zhu AX, Finn RS, Edeline J, et al. Pembrolizumab in patients with advanced hepatocellular carcinoma previously treated with sorafenib (KEYNOTE-224): a non-randomised, open-label phase 2 trial. Lancet Oncol. 2018;19(7):940-952. doi: 10.1016/S1470-2045(18)30351-30356.