Design, Criteria, and Endpoints of KEYNOTE 564


Nizar Tannir, MD, describes the data, criteria, and end points of the KEYNOTE-564 trial.

Nizar Tannir, MD: Here’s the design of KEYNOTE-564. The key eligibility criteria included confirmed clear cell RCC [renal cell carcinoma], nephrectomy within 12 weeks prior to randomization—that’s important—good performance status, and stratification factor M0 vs M1 NED [no evidence of disease]. This trial recruited patients who had metastatic disease, oligometastases that were resected to no evidence of disease status. That’s the M1 NED. That’s one stratum. The other one within the M0 group was further stratified by ECOG 0 vs ECOG 1 and US vs non-US sites. The experimental arm in KEYNOTE-564 was pembrolizumab at 200 mg IV [intravenously] every 3 weeks for 1 year vs placebo for 1 year. The primary end point was disease-free survival [DFS] by investigator. The key secondary end point was OS [overall survival]. The other secondary end point was safety. In the next few slides, we’ll see the efficacy results from KEYNOTE-564.

Here are the patients who were eligible. There were 3 cohorts of patients. The vast majority of those patients were in the intermediate-high [risk cohort], and fewer patients—5% to 8%—were M1 NED and had high risk as defined by pT4, or any tumor—any pT—any grade, but node positive. Those plus the M1 NED constituted about 13% of the recruitment, and the vast majority of patients were intermediate-high [risk], pT2, grade 4, or sarcomatoid, N0, M0, and then any grade but pT3, N0, M0.

At the bottom are data for disease-free survival at 5 years by UISS [UCLA Integrated Staging System] and for the M1 NED, the ECOG 2810 study. The probability of relapse is highest for M1 NED, and next is the node-positive patients. Next is patients who have T4 disease, and then pT3 disease, which this patient presented with. That patient will have a relapse rate of 20% to 45%. The pT3 [category] comes in different flavors: pT3a, pT3b, or pT3c. The higher the grade, the higher the probability of relapse. That speaks to why I’m not sending those patients to Scott [Tykodi] or Moshe [Ornstein] at their institutions. Those patients have only 20% probability of relapse at 5 years if they have pT2 disease, grade 4.

The list of the criteria is on the slide. As I mentioned, 86% or 87% of those patients had M0, intermediate-high risk. Those were the T2, grade 4, or sarcomatoid and T3 disease. Eighty-five percent of the M0 high-risk [patients] were M1 NED.

[Let’s look at] the couple of markers for the primary end point, disease-free survival. This is intent to treat by investigator assessment. The green curve represents patients treated with the pembrolizumab, and the red [curve] is patients treated with placebo. At the 2-year mark, there’s about an 11% delta for disease-free survival probability between the 2 arms. That had a hazard ratio of 0.68 at the time of the initial analysis at ASCO [American Society of Clinical Oncology Annual Meeting] 2021. That was a statistically significant difference for DFS between the 2 arms. At subsequent follow-up with a minimum follow-up of 2½ years, that 11% difference is maintained, with a hazard ratio a little lower, at 0.63. [This is] a significant difference for the delta for disease-free survival probability at 2 years.

Looking by risk category—intermediate-high risk, high risk, and M1 NED—for all 3 cohorts, there’s a separation of the curves for disease-free survival between the treatment arm, pembrolizumab vs placebo, with hazard ratios of 0.68, 0.60, and 0.28. But a cautioning note here: for the high risk and the M1 NED, because the number of patients is small, I’d interpret that with caution. If you look at the confidence interval for high risk, the upper bound is 1.10. It crosses the 1. The hazard ratio is 0.6. The difference is numerically higher, but it’s a small number of patients, so one has to be careful in interpreting this.

When you have an adjuvant therapy and you haven’t shown significant improvement in OS because the OS data are still immature, safety becomes important. One will look at the subgroups for trends or hints. You can look at the forest plot and argue about whether forest plots are useful. When you look at the randomized phase 3 trials, you can see patients treated with pembrolizumab across all those categories of subgroups, including age, sex, ECOG [performance status], region, and tumor grade. PD-L1 status favored pembrolizumab over placebo, although one has to be careful about putting too much emphasis on subgroup analysis with forest plots from these trials.

This is the interim OS in the intent-to-treat population. The primary analysis had a median follow-up of 24.1 months on the left, and [the updated analysis] had a 30.1-month median follow-up. The top curve, the green curve, represents patients treated with pembrolizumab; the red curve represents patients treated with placebo. The hazard ratio is 0.54, but only a third of the deaths have been observed. These are still immature data for OS. We need longer follow-up to say whether this will cross the prespecified statistical significance. However, it’s important to know that there have been more deaths on the placebo arm. You can see patients with events at the latest follow-up: 43 vs 23. There’s a difference there, but the median hasn’t been reached for either arm.

[Looking at] the adverse events [AEs], if we focus on the [updated analysis], there are higher [rates of] all-cause AE in the pembrolizumab arm vs the placebo arm—especially when you look at grade 3 and higher adverse events, which were almost double with pembrolizumab over placebo. That’s no surprise. But there were only 2 deaths in the pembrolizumab arm and 1 death in the placebo arm. Those 2 deaths aren’t necessarily from the study. They weren’t necessarily treatment related. The serious AEs were almost double with pembrolizumab vs placebo, and serious AEs leading to treatment discontinuation were 10% vs 1%, which wasn’t a surprise. Even in the healthy, fit population, such as patients postnephrectomy, there are patients who develop toxicity from an anti–PD-1, which leads to discontinuation. Grade 3 and 4 treatment-related adverse events were 18.6% vs 1.2%. Immune-related adverse events were lower: about 9% vs 0.6%. Around 8% of the patients required high-dose corticosteroids for management of immune-related adverse events.

[This slide] talks a little more about the adverse events. We’re all familiar with adverse events from pembrolizumab monotherapy. Side by side in that tornado plot are patients treated with pembrolizumab and patients treated with placebo. Top to bottom are adverse events we expect from an anti–PD-1: fatigue, rash, pruritus, thyroid disorder, diarrhea, arthralgia, nausea, myalgia, and asthenia. The median time to first onset of these adverse events in the pembrolizumab arm is about 6 months.

Transcript edited for clarity.

Related Videos
Video 11 - "Treating Patients With RCC Who Progress Following Adjuvant Therapy"
Video 10 - "RCC: Informing Treatment Decisions with Clinical Trial Data"
Video 9 - "KEYNOTE-564: Adjuvant Pembrolizumab in Renal Cell Carcinoma"
Video 8 - "Adjuvant Immunotherapy in Renal Cell Carcinoma"
Video 7 - "Case 2: A 48-Year-Old Man with Renal Cell Carcinoma"
Related Content