A PERFECT Biomarker-focused Study of Neoadjuvant IO for Esophagogastric Cancer

In this issue of Clinical Cancer Research, van den Ende and colleagues assessed the feasibility and preliminary efficacy of 40 patients treated with neoadjuvant chemoradiotherapy (nCRT) with programmed-death ligand-1 (PD-L1) immune checkpoint inhibition (ICI) using atezolizumab for resectable locally advanced esophagus and gastroesophageal junction adenocarcinoma (1). This pilot, single-arm, open-label, “PERFECT” study entailed sequential CROSS nCRT (carboplatin/paclitaxel with 41.4 Gy radiation, 23 fractions) weeks 1–5 combined with atezolizumab (weeks 1 and 4), then atezolizumab alone for three cycles (weeks 7, 10, and 13), and then surgery (between weeks 14 and 16). The primary endpoint was feasibility and safety, defined as ≥75% patients successfully receiving all five planned atezolizumab cycles; this criterion was met with 85% of patients completing the atezolizumab. However, autoimmune adverse events were experienced in 15% of patients, 33% of patients died or were hospitalized during the neoadjuvant therapy, and 12.5% of the patients experienced disease progression or death prior to surgery. Also, 83% of patients had surgical resection, lower than historical resection rates of 89% with nCRT or 94% with chemotherapy (2). While efficacy endpoints of nCRT+ICI were exploratory, the authors compared observed outcomes with propensity-matched patients from the Netherlands Cancer Registry. They concluded that the complete pathologic treatment response grade-1 (TRG1) rates of 10/40 (25%) versus 27/134 (20.1%), respectively, were not significantly different. Likewise, while overall survival (OS) and disease-free survival (DFS) were not significantly different, they, in fact, actually trended to worse in the not-so-PERFECT cohort. Important exploratory translational work evaluating the impact of nCRT+ICI on the tumor immune microenvironment (TIME) was performed at baseline, on-treatment (week 3), and on surgical specimens when available. These analyses demonstrated that pathologic responses (TRG1+TRG2), totaling 37.5%, were enriched only in patients with baseline inflamed TIME prior to nCRT, as measured by high PD-L1 combined positivity scores (CPS) ≥25 and high IFNγ signatures. In contrast, nonresponders demonstrated two biological subgroups. One subgroup demonstrated no/low CTLs before or during therapy. The other demonstrated modulation of the TIME during nCRT showing CTLs within the tumor bed with upregulation of various checkpoints. This inflammatory surge observed only after nCRT did not seem to predict any therapeutic benefit. These are important findings if validated in larger independent series.

The timing of this publication is impeccable, being the first to prospectively assess ICIs in the neoadjuvant setting of this disease. With demonstrated benefit of anti-PD-1/PD-L1 ICIs in the metastatic setting for PDL1-highly expressing gastroesophageal adenocarcinomas (GEA), the natural course is to now investigate in the locally advanced setting, with objectives to enhance long-term DFS and OS. Currently, the perioperative treatment landscape includes two approaches for locally advanced esophagogastric adenocarcinomas, including CROSS nCRT versus perioperative triplet chemotherapy with FLOT (5-fluorouracil, leucovorin, oxaliplatin, and docetaxel; ref. 2). An ongoing head-to-head study (ESOPEC, NCT02509286) to address the debate as to which, if any, approach is superior as a backbone therapy, prior to even considering the addition of ICIs, is now fully accrued and awaiting data maturity. Hypothetically, either chemotherapy or nCRT could modulate the TIME to enhance the benefits of ICI therapy (3, 4). Therefore, the follow-up question going forward, after observing the results of ESOPEC, is whether or not there are differences between these two backbone strategies pertaining to modulating the TIME to potentiate the activity of ICIs. Notably, an exploratory study of 31 patients evaluating the impact of nCRT on TIME revealed that nCRT increased PD-L1 expression from 14 of 31 to 24 of 31 patients (+32%), and also upregulated various other checkpoints, CD8⁺ cells, and IFNγ signature (4), observations recapitulated here by PERFECT. Relevantly in that study, there were no appreciable differences in TRG in those inflamed versus noninflamed tumors at baseline prior to, nor between those tumors becoming inflamed or not after nCRT alone (4). Those observations suggest that inflammation before or during/after nCRT is not predictive of benefit from nCRT alone. If confirmed, these are important considerations as historical controls for the current nonrandomized study now using nCRT+ICI. Here, the benefit of atezolizumab was limited to those tumors with baseline inflammation, most notably in PDL1 CPS≥25 tumors, which demonstrated a TRG1+2 of 54% (7/13), versus 26% (6/23) in those tumors with CPS < 25. A TRG1+2 rate of 26% is one that would be expected (or lower than expected) with nCRT or chemotherapy alone (2). The incidence of CPS ≥ 25 was 33% (13/36) of patients within the analysis, and this was the only group having improved OS and DFS compared with the propensity-matched patients.

It is important to consider results of this PERFECT study in context of the recently reported randomized phase III adjuvant Checkmate-577 study (5). Checkmate-577 evaluated adjuvant nivolumab for 1 year versus placebo in patients with esophageal and gastroesophageal junction cancer (adenocarcinoma and squamous cell). After nCRT and resection with clear margins, those with residual disease in the surgical specimen, particularly those with node positive residual disease, have a relatively poor prognosis. In the first report of Checkmate-577 evaluating these patients, median DFS was improved from 11.0 to 22.4 months, and estimated 3-year DFS rates were approximately 40%–45% versus approximately 30%–35% with nivolumab versus placebo, respectively, portending an absolute benefit of approximately 10%–15%. This absolute benefit is similar to that seen in metastatic gastroesophageal adenocarcinoma settings both with ICI monotherapy or with chemotherapy. Checkmate-577 results do require longer follow-up to more accurately assess DFS and also OS, but most importantly, must discern whether there is differential benefit by histology by PD-L1 CPS scoring, which currently remains imperfectly reported. In PERFECT, an enhanced long-term DFS and OS among responders to nCRT + ICI compared with the propensity-matched patients was observed only in patients with inflamed tumors, which at this point should not be surprising to anyone. That the benefit was only observed in baseline inflamed tumors and not those with induced inflammation only during/after therapy is intriguing, and will be important to confirm in other prospective independent analyses, as this finding would have very significant implications.

Targeted therapies, including ICIs, for metastatic gastroesophageal adenocarcinoma have a recurring theme that significant benefit is realized only within biomarker-select subgroups—and for ICIs that is high levels of PD-L1 expression by CPS scoring. The importance of PERFECT, in addition to demonstrating relative feasibility in the neoadjuvant setting, is that it is the first to report efficacy which yet again depends on this same biomarker. Studies evaluating the utility of ICI therapy should therefore be required to transparently report outcomes by PD-L1 CPS scoring at several cutoffs (including lower expressing groups explicitly), to continue to determine which patients actually derive benefit. This is of particular importance now since some already consider 1 year of adjuvant nivolumab “standard,” per Checkmate-577 criteria, yet without this detailed biomarker information reported. What has now been reported is that CPS < 5 tumors (combining both adenocarcinoma and squamous cell) did not derive significant benefit (N = 295; HR 0.89, 95% CI 0.65–1.22) (5). It would be expected that PD-L1 low/negative adenocarcinomas would demonstrate even worse outcomes compared to squamous cell. This histologic biomarker subgroup data then would be critical to explicitly report and consider. Sparing patients from ineffective neaodjuvant therapy and/or year-long adjuvant therapy will be important to prevent unnecessarily increased clinical and financial toxicities. Furthermore, the authors astutely point out that 1 year of ICI therapy adjuvantly may not be necessary, surmising that short-course preoperative therapy could sufficiently demonstrate the same benefit in those destined to derive it. Larger studies should be performed to confirm these PERFECT observations, and seek to optimize all iterations of the type, sequence, and length of therapies including chemotherapy, chemoradiotherapy, and ICIs. Future therapeutic strategies could entail prospective stratification of inflamed versus noninflamed tumors both at baseline and on treatment, and assign immune-based therapies accordingly (Fig. 1).

As such, PERFECT serves as a benchmark of perioperative ICI-directed therapies for gastroesophageal adenocarcinoma upon which to build. A number of ongoing studies are evaluating ICI therapy pre- and/or post-surgery either as monotherapy or concurrently with chemotherapy, chemoradiotherapy, and/or other ICIs. These include EA2174 (NCT03604991), KEYNOTE-585 (NCT03221426), and MATTERHORN (NCT04592913), among others. These studies should be able to confirm these first observations of feasibility, safety, and efficacy. Hopefully those studies too will perform the necessary biomarker analyses to narrow down the population gaining benefit from ICIs to optimize value and justify risks of perioperative ICIs. If they do, well, that would be just…PERFECT.

D.V.T. Catenacci reports personal fees from Merck, BMS, Lilly, Taiho, Five Prime, Astellas, Genentech/Roche, Amgen, Seattle Genetics, Daiichi Sankyo, QED, Basilea, Tempus, Guardant Health, Natera, and Archer during the conduct of the study.

D.V.T. Catenacci is supported by the NCI (P30CA014599).