Results and Molecular Correlates from a Pilot Study of Neoadjuvant Induction FOLFIRINOX Followed by Chemoradiation and Surgery for Gastroesophageal Adenocarcinomas

Despite recent advancements, gastric and gastroesophageal junction (GEJ) adenocarcinoma remains a lethal disease (1, 2). While surgical resection offers the potential for cure in locoregional disease, the majority of patients experience disease recurrence within 5 years (3). To that end, numerous treatment paradigms and regimens have been explored to improve upon the outcomes of surgery alone (3–9). For gastric cancer, perioperative chemotherapy and postoperative chemotherapy have improved outcomes compared with surgery alone (4–8). For locally advanced GEJ cancer, preoperative chemoradiation and perioperative chemotherapy have been established as standards of care (4, 9). However, despite these recent advancements, the risk of distant recurrence remains high, at over 50% by 5 years (3, 5). Moreover, randomized data of perioperative treatment approaches have demonstrated the challenges of administration of postoperative therapy, with treatment completion rates of approximately 40%–50% (4–5). Therefore, in this current study, we designed a total neoadjuvant treatment strategy with the goal of further improving treatment compliance and potentially, outcomes. In addition, strategies to optimize the integration and sequencing of multimodality therapy, coupled to biomarkers of response, also remain a significant unmet need (10–12).

Outside of gastroesophageal adenocarcinoma, FOLFIRINOX has emerged as a standard of care for pancreatic and colorectal cancer in both the resectable and metastatic setting (13–16). In addition, there is a growing body of literature evaluating the efficacy of this regimen in gastroesophageal and gastric cancers (17–19). A pooled analysis of two ongoing phase II trials of FOLFIRINOX in the first-line metastatic setting for GEJ cancers demonstrated an overall response rate (ORR) of 62.5% (17). In a recently published phase II trial of advanced gastroesophageal adenocarcinoma, among ERBB2-negative patients treated with FOLFIRINOX alone, the ORR was 61%, with a median overall survival (OS) of 15.5 months. Among ERBB2-positive patients treated with FOLFIRINOX and trastuzumab, the ORR was 85%, with a median OS 19.6 months (18). In addition, a multi-institutional phase II trial of 36 patients with locally advanced gastroesophageal adenocarcinoma evaluated pharmacogenomically dosed perioperative FOLFIRINOX and reported R0 resection rate of 92% and a pathologic complete response (pCR) rate of 11% (19). Given the high rates of efficacy seen in pancreatic and colorectal cancer (13–16), the familiarity of our providers in administration and management of treatment-related toxicities of this 3-drug regimen, and the promising results appreciated in preliminary studies discussed above (17–19), we sought to perform a single-arm pilot study of total neoadjuvant therapy with FOLFIRINOX, followed by consolidative chemoradiation with concurrent carboplatin/paclitaxel, followed by surgery, in patients with gastroesophageal adenocarcinoma planned for curative intent therapy.

Patients with previously untreated gastric or GEJ adenocarcinoma were enrolled in this NCI-sponsored single-arm pilot clinical trial from October 2017 to June 2018 (NCT03279237). This study was approved by the Dana-Farber Cancer Institute/Harvard Cancer Center (Boston, MA) Institutional Review Board. All patients provided written informed consent. Prior to study enrollment, all patients underwent the following evaluation: (i) clinical staging, including esophagogastroduodenoscopy (EGD) ± endoscopic ultrasound, CT scan of the chest and either a CT scan of the abdomen/pelvis or a MRI scan of the abdomen with intravenous contrast within 42 days of enrollment, (ii) physical examination within 14 days of enrollment, and (iii) baseline laboratory studies, including urine HCG for women of childbearing potential. Microsatellite instability (MSI) status was not routinely performed. Staging laparoscopy was not required prior to study treatment and was at the discretion of the treating investigator.

Study eligibility criteria were as follows: histologically or cytologically confirmed T3/4 or lymph node (LN)-positive disease (defined as > 1 cm in size and/or FDG-avid), centralized pathologic confirmation; age ≥18 years; Eastern Cooperative Oncology Group performance status (ECOG PS) ≤ 1; life expectancy > 3 months; and adequate organ and marrow function as defined as absolute neutrophil count ≥ 1,500 cells/mm³; platelets ≥ 75,000 cells/mm³; aspartate aminotransferase (SGOT) and alanine aminotransferase (SGPT) ≤ 2.5 × upper limit of normal, and creatinine ≤ 1.5 mg/dL, or creatinine clearance ≥ 30 mL/minute/1.73 m² for participants with creatinine levels above institutional normal; and the ability to understand and the willingness to sign a written informed consent document. The full protocol is provided in supplemental data (Supplementary Appendix S1).

Exclusion criteria included: evidence of metastatic disease within 6 weeks of study entry, prior treatment for the participant's gastric or GEJ cancer; treatment of other invasive carcinomas within the last 5 years with greater than 5% risk of recurrence at time of eligibility screening; receipt of any other investigational agents within 4 weeks preceding the start of study treatment; serious concomitant systemic disorders; pregnancy; major surgery (excluding laparoscopy) within 4 weeks of start of study treatment; prior systemic fluoropyrimidine therapy or known hypersensitivity to 5-fluorouracil (5-FU) or known dihydropyrimidine dehydrogenase deficiency; and a history of allergic reaction(s) attributed to compounds of similar chemical or biologic composition to 5-FU, irinotecan, or oxaliplatin. Distant nodal disease was allowed if it was deemed to be covered within a radiation treatment field and curative therapy was planned.

This is a single-arm pilot study of the neoadjuvant administration of the FOLFIRINOX regimen and preoperative radiation therapy followed by surgery in patients with locally advanced (T 3/4 or N+) gastric or gastroesophageal adenocarcinoma. The primary study objective was to determine the completion rate of neoadjuvant FOLFIRINOX followed by consolidative chemoradiation with concurrent carboplatin/paclitaxel delivered in the preoperative setting. Secondary objectives included: (i) acute toxicity; (ii) pCR rate; (iii) progression-free survival (PFS); (iv) OS. Pre-planned correlative studies to explore the relationship between circulating tumor DNA (ctDNA) and clinical outcomes were performed. Exploratory analyses were performed to evaluate tumor microenvironment and molecular profiling.

FOLFIRINOX was planned for eight cycles on a 14-day cycle. 5-FU was administered as 400 mg/m² bolus on day 1, then as a 2,400 mg/m² continuous infusion for 46 hours. Leucovorin calcium, 400 mg/m²; oxaliplatin 85 mg/m², and irinotecan hydrochloride 180 mg/m², were administered on day 1, as described previously (14). Pegfilgrastim, 6 mg, was administered on day 4. Dose adjustments for toxicity were predefined in the protocol (Supplementary Appendix S2). Patients were restaged with CT scans of the chest, abdomen, and pelvis after four and eight cycles of FOLFIRINOX.

Chemoradiation treatments were scheduled to begin within 4 weeks after completion of FOLFIRINOX. Carboplatin and paclitaxel were administered concurrently with the radiotherapy. Carboplatin (AUC = 2) was administered weekly over 1 hour. Paclitaxel 50 mg/m² was administered as an intravenous infusion over 30–60 minutes weekly as described previously (9).

All radiation treatments were administered at the Clark Center for Radiation Oncology at Massachusetts General Hospital (Boston, MA). All patients underwent 4D-CT simulation with oral and intravenous contrast in a supine position. For photons, all patients received Intensity Modulated Radiation Therapy or Volumetric Modulated Arc Therapy. Daily cone beam CT was performed pre-treatment for setup verification. Tumor volume was defined on the basis of CT, PET/CT, and MRI. The gross tumor volume (GTV) was defined as the gross primary tumor and any LNs ≥1 cm. For participants with diffuse type gastric cancer, where the GTV could not be visualized, the entire stomach was identified as GTV. When the GTV was clearly identified, the clinical target volume (CTV) included a longitudinal mucosal margin of 3.5–4 cm proximally and distally to the GTV. A 5 mm CTV expansion was generated on the basis of all grossly enlarged LNs. Elective LN coverage included celiac and gastrohepatic LNs for all patients. For GEJ and cardia lesions, coverage of paraesophageal LNs within 4 cm of the tumor was recommended; for distal gastric lesions, porta hepatis LN coverage was recommended. Coverage of splenic LNs (for GEJ/cardia lesions) and peripancreatic LNs (for distal gastric lesions) were at the discretion of the treating physician. The planning target volume (PTV) was customized on the basis of the 4D CT scan; however, generally, a 5 mm expansion was used, except for 7 mm superiorly/inferiorly. The prescribed dose to the PTV for gastric tumors was 4,500 cGy delivered in 180 cGy/day over 25 fractions. For GEJ tumors, a 540 cGy boost was included to the GTV + 1 cm. Normal dose constraints were predefined in the protocol (Supplementary Appendix S3).

After completion of chemoradiation, restaging CT scan was evaluated by the multidisciplinary team. Surgical resection was performed approximately 4–6 weeks after completion of chemoradiation. Surgical procedure and extent of LN dissection was performed at the discretion of the treating surgeon. For gastric cancers, an extended (D1+ or D2) LN dissection was performed.

After treatment completion, study participants had a follow-up baseline CT scan performed 3–8 weeks postoperatively. Subsequently, follow-up visits were scheduled with laboratory evaluations every 3 months and with CT scans every 6 months for the first 2 years, visits with blood work every 3 months and annual CT scans for year 3, and visits with blood work at least every 6 months and annual CT scans up to year 5. Routine surveillance EGD was not performed.

Pathologic findings were scored per standard institutional practices, including margin status (proximal, distal, radial) and nodal status (total assessed, total positive). When applicable, pathology reports specified presence of pathologic findings consistent with nodal clearance (i.e., acellular mucin, fibrosis). Tumor regression grade (TRG; ref. 20) was recorded for all patients undergoing surgery, and pathologic complete response was defined as TRG 0.

Tumor biopsies from the time of diagnosis as part of routine care were subjected to an internal next-generation sequencing (NGS) platform evaluating 104 known cancer genes. Blood was drawn at prespecified study timepoints (baseline, cycle 1 day 8, cycle 5 day 1, pre-chemoradiation, preoperatively, every 3 months postoperatively, and at progression) at time of draws for routine clinical care. ctDNA was extracted from plasma using QIAGEN-based protocols and amplified by digital droplet PCR (ddPCR) using primers for tumor-specific point mutations. One or more driver mutations that matched the patient's tumor sequencing were followed longitudinally.

Molecular profiling of hotspot SNP, indel, and copy-number variants was performed using SNaPshot analysis, an Anchored Multiplex PCR strategy described previously (21). Briefly, after pathologic review to determine sample tumor enrichment, genomic DNA was isolated from formalin-fixed paraffin-embedded tumor specimens. Genomic DNA was sheared, end-repaired, A-tailed and then modified by ligation of half-functional Illumina adapters. Two hemi-nested PCR reactions targeting genomic regions of interest (i.e., SNP/mutation, indel, or copy-number hotspots; Supplementary Appendix S4) were performed, and sequenced using Illumina Nextseq 2×150 bp paired-end sequencing. Genome alignment to the hg19 reference sequence was performed using Novoalign (Novocraft Technologies) and various methods for variant calls were employed, including use of MuTect1 (22), LoFreq (23), GATK (24–26) and a laboratory-developed hotspot and copy-number caller. This test is validated to detect variants with at least 5% allelic frequency, in regions with sufficient NGS read coverage.

Data from The Cancer Genome Atlas (TCGA; ref. 27) were analyzed using cBioportal (28) and Xenabrowser.net (29) to assess co-mutation frequency and alteration subtype specificity.

The primary endpoint was neoadjuvant therapy completion, defined as the portion of patients receiving all planned induction FOLFIRINOX followed by chemoradiation. Sample size was planned for 25 patients, and if at least 18 of the 25 patients were able to complete the specified treatment plan, then the primary endpoint was met. The decision rule was associated with 89% power for declaring success if FOLFIRINOX in combination with chemoradiation were associated with an underlying completion rate of 80%. In contrast, the probability of a type I error was only 15% if 60% of patients were able to complete the intended treatment plan. Toxicity was defined per CTCAE v4.03. PFS and OS were measured from the date of treatment start and estimated by the Kaplan–Meier (KM) method. OS time was censored at the date of last follow-up for patients still alive. PFS time was defined until detection of locoregional recurrence, distant metastases, or death without documented progression, whichever date was earliest, or censored at the date of last follow-up. The distant metastasis (DM) rate was estimated as the cumulative incidence with progression or death before chemoradiation as a competing risk. Fisher exact test and Gray test were used, respectively, to analyze pCR and DM rates by ctDNA detection status, with the P values based on a two-sided hypothesis. As the biomarker analysis was exploratory, the level of type 1 error was not pre-specified. Statistical analysis was performed using SAS version 9.4 (SAS Institute) and R version 3.3.1 (R Foundation).

Between October 2017 and June 2018, a total of 25 eligible patients were enrolled. Table 1 describes baseline patient and tumor characteristics. The median age was 60 (range, 30–76), and 68% were male. The primary site was 60% GEJ and 40% stomach. The clinical stage of disease was T3N0 (8%), T4N0 (12%), T1–4N+ (80%). At time of presentation, 28% of patients had nodal M1 disease including supraclavicular (n = 4), retroperitoneal (n = 4), mediastinal (n = 3), and retrocrural (n = 1) LN. Four patients had multiple sites of distant LN involvement. Eight patients (32%) underwent diagnostic laparoscopy.

Of 25 patients who started treatment, 2 patients developed disease progression on FOLFIRINOX. One patient developed a liver metastasis on first set of restaging scans, and a second patient with a GEJ primary lesion developed gastric outlet obstruction from a distinct gastric lesion in the pylorus. Of the remaining 23 patients who completed FOLFIRINOX, 23 of 23 (100%) proceeded to chemoradiation. One patient died during chemoradiation of a cardiac pulseless electrical activity (PEA) arrest, which was deemed unrelated to treatment. Of the remaining 22 patients (88%) who completed chemoradiation, all proceeded to surgical exploration. Thus, 22 of 25 (88%) enrolled patients completed the induction FOLFIRINOX and chemoradiotherapy. During surgery, 2 of the 22 patients were found with intraoperative metastases (1 patient with peritoneal disease, 1 patient with peritoneal and liver disease), neither of whom underwent pre-treatment diagnostic laparoscopy. Therefore, a total of 20 patients (80%) underwent potentially curative surgical resection (Fig. 1). Among the 4 patients who had disease progression over the course of treatment, none had nodal M1 disease at presentation.

Of the 20 patients that went to resection, the median residual tumor size was 0.25 cm (range, 0.0–7.7 cm). Six (30%) patients had positive LNs, and all patients had a R0 resection [intention to treat (ITT) R0 resection: 80%]. Surgical outcomes are summarized in Table 2. Seven patients achieved a pCR, corresponding to 35% of the resected patients and 28% of the entire cohort by ITT.

The median follow-up of the 18 surviving patients was 23.1 months (range, 19.2–26.9 months). Among the 20 patients who underwent resection, 14 are still alive without evidence of relapse. In the entire cohort, the 1-year DM rate, PFS, and OS were 27%, 64%, and 76%, respectively; the 2-year DM, PFS, and OS rate were 37%, 55%, and 72%, respectively (Fig. 2). Excluding the 7 patients with nodal M1 disease, the 1-year and 2-year DM rate, PFS and OS were 25% and 39%, 67% and 54%, and 78% and 72%, respectively. Among the 8 patients with progression following chemoradiation, the site of first failure included the liver (n = 3), peritoneum (n = 3), lung (n = 1), and brain (n = 2).

Among patients with nodal M1 disease (n = 7), all patients completed eight cycles of FOLFIRINOX with radiographic response, and all but 1 of the patients went for surgical resection. The only patient with nodal M1 disease who did not undergo surgical resection died unexpectedly during chemoradiation. Of the remaining 6 patients, 2 have subsequently developed metastases in the brain and liver, respectively, and 4 patients remain disease free.

Supplementary Appendix S5 summarizes the acute toxicities of this treatment regimen. The overall rates of acute grade 3 and 4 toxicity were 16% and 80%, respectively. However, in all but 2 patients (8%), grade 4 toxicity was secondary to subclinical lymphopenia during chemoradiation. To that end, the rate of non-hematologic grade 3+ and grade 4+ worst toxicity was 48% and 0%, respectively. As mentioned above, one patient died during the course of chemoradiation at 46.8 Gy from cardiac arrest (PEA arrest), which was clinically most consistent with pulmonary embolus and felt to be unrelated to treatment.

All patients had tissue available for tumor-specific mutational analysis. Across all patients, tumor-specific mutations were identified in 21 (84%) patients by ctDNA ddPCR. All but one of the patients were tested by ddPCR for 1 mutation, and 1 patient was tested for two mutations. Thus, the range of ddPCR mutations is 1–2, and the median = 1. Of the 21 patients with an evaluable mutation for ddPCR (Supplementary Table S6), 4 patients had undetectable ctDNA at baseline and 17 had detectable ctDNA pre-treatment. Undetectable ctDNA was associated consistently with a numerically higher pCR rate (∼50%) across all timepoints compared with detectable ctDNA, although this was not statistically significant (Table 3).

A timeline of ctDNA for all patients with evaluable ctDNA is provided in Fig. 3. Among the 16 resected patients with post-chemoradiation ctDNA draws, only 1 of the 12 patients with undetectable ctDNA (8%) developed disease recurrence, compared with 3 of the 4 patients with detectable ctDNA (75%; P = 0.004). Among the 14 resected patients with postoperative ctDNA draws, none of the 9 patients with undetectable ctDNA (0%) developed recurrence versus 2 of 5 patients (40%) with detectable ctDNA (P = 0.045) after surgery.

Twenty pre-treatment and 13 post-treatment samples were analyzed by SNaPshot molecular profiling, with paired pre- and post-treatment analysis available for 11 patients (Supplemental Appendix S7). Review of variants in pre- and post-treatment samples revealed 3 patients with evidence of potential treatment-induced subclonal selection and/or tumor evolution (Fig. 4). The pre-treatment sample for Patient 25 with a cancer of the gastric antrum demonstrated a KRAS G13D mutation, and this patient was known to have MSI as a consequence of MLH1/PMS2 methylation-based gene repression, consistent with MSI-high gastric cancer. Post-treatment molecular analysis revealed eight additional MSI subtype-related mutations, consistent with a hypermutable state (Fig. 4A).

The pre-treatment sample for Patient 24 with GEJ cancer possessed molecular variants in TP53, RNF43, AKT1, and SMARCA4. Post-treatment analysis (tumor purity 60%) detected new PIK3CA (three distinct mutations), MAP3K1, MYCN, and FGFR2 mutations but did not detect the TP53 or RNF43 variants, suggestive of loss of subclone(s) possessing TP53 and RNF alterations (Fig. 4B). Numerous PIK3CA mutations were detected in the post-treatment sample, which could reflect: (i) single clone with biallelic H1047 mutations, at least one allele of which co-occurs with a C420R mutation, or (ii) multiple clones with different PIK3CA mutation permutations. Of note, the epigenetic disruption from SMARCA4 mutations has been previously reported to result in gastric tumor subclonal subtype plasticity (30), which may be reflected in the numerous molecular variants evident on post-treatment sequencing.

Patient 19 with a GEJ adenocarcinoma demonstrated no variants on pre-treatment sample analysis (tumor purity 70%), while post-treatment analysis indicated four molecular variants (TP53, KRAS amplification, PIK3CA, and FBXW7). Analysis of TCGA gastric tumor data (Supplementary Appendix S8) revealed that none of the PIK3CA-mutant TCGA gastric tumors possessed KRAS amplifications, and vice versa. In addition, KRAS amplified and TP53 mutations clustered in CIN subtype tumors, PIK3CA and FBXW7 mutations were associated with MSI tumors, and co-mutational statuses were even more specific for these two subtypes. Assessing Patient 19's tumor evolution using this information suggests two models of tumor evolution (Fig. 4C): (i) typical “accumulation” of mutations, and (ii) establishment and outgrowth of two putative subclones of different subtypes.

There are currently multiple different treatment paradigms, all supported by randomized controlled trials, which are deemed acceptable as standard-of-care treatment for locally advanced GEJ and gastric cancers (4–9). For GEJ cancers, the CROSS study established neoadjuvant chemoradiation with concurrent carboplatin/paclitaxel as one acceptable standard of care for treatment of locally advanced distal esophageal and GEJ cancers, yielding improved OS compared with surgery alone (9). Perioperative chemotherapy alone is another acceptable treatment for GEJ and gastric cancers, having demonstrated improved outcomes compared with surgery alone (5). More recently, the FLOT-4 study established the superiority of perioperative FLOT (5-FU, docetaxel, and oxaliplatin) to perioperative ECF (epirubicin, cisplatin, and 5-FU; ref. 4). However, perioperative approaches are plagued by low rates of receipt of postoperative therapy, potentially limiting efficacy (3–6). Therefore, in this current study, we designed a total neoadjuvant treatment strategy with the goal of further improving treatment tolerability, compliance and potentially, outcomes. Our study mimics the movement in currently ongoing randomized trials in this space, including the CRITICS-II (10) and TOPGEAR (11) studies, which frontload chemotherapy and chemoradiation prior to surgical resection.

Despite these promising advancements in the treatment of gastric and gastroesophageal adenocarcinoma cancers, the 5-year OS is still only in the range of 40%–50% (3–6, 9). The majority of patients with recurrence will subsequently die of metastatic disease, suggesting a need for systemic intensification to reduce recurrence rates and to prolong survival (3, 5, 31). In designing our current study, given the effectiveness of this regimen in extending survival in pancreatic cancer and colorectal cancer (13–16), as well as our relative fluency in administering it and managing treatment-related toxicities, we decided to proceed with FOLFIRINOX as a chemotherapy backbone. To date however, there has been little published data regarding its efficacy for gastroesophageal and gastric cancers. In 2017, a pooled analysis of two ongoing phase II trials of FOLFIRINOX in the first-line metastatic setting for GEJ cancers demonstrated an overall response rate of 62.5% (17). More recently, in a phase II trial from Washington University (NCT 01928290), among ERBB2-negative patients treated with FOLFIRINOX alone, the ORR was 61%, with a median OS of 15.5 months. Among ERBB2-positive patients treated with FOLFIRINOX and trastuzumab, the ORR was 85%, with a median OS 19.6 months, (18) which offers promising results compared with historical controls (32). In addition, a multi-institutional phase II trial of 36 patients with locally advanced adenocarcinoma evaluated pharmacogenomically dosed perioperative FOLFIRINOX and reported R0 resection rate of 92% and a pCR rate of 11% (19).

Building on the high response rates seen in the above mentioned studies (17–19), we performed a pilot study of neoadjuvant FOLFIRINOX followed by consolidative chemoradiation with carboplatin/paclitaxel. We determined that completion of this total neoadjuvant regimen was feasible, with an 88% therapy completion rate, successfully meeting our primary endpoint. Although we noted a grade 4 toxicity rate of 80%, the majority of grade 4 toxicity was attributable to subclinical lymphopenia (76%) during induction FOLFIRINOX without clinical implications. The remaining grade 4 toxicity was attributable to only 2 patients (8%) with grade 4 febrile neutropenia. Overall, the toxicity profile is consistent with that reported in the literature for the various treatment components (33). Given the small numbers and heterogeneity in disease sites, it is difficult to cross compare between studies with respect to acute toxicity. Perioperative FLOT had 51% grade 3+ neutropenia (4), compared with 8% in our current study, which likely reflects the routine practice of administration of pegfilgrastim, 6 mg, on day 4 of each cycle. Compared with perioperative FLOT (4) and neoadjuvant CROSS regimens (9), our current regimen reported higher rates of grade 3+ thrombocytopenia (16% vs. perioperative FLOT 2% vs. CROSS 1%). Grade 3+ diarrhea may be more prevalent in our current regimen (16%) and perioperative FLOT (ref. 4; 10%) compared with CROSS (ref. 9; 1%) due to the impact of 5-FU the former paradigms, and irinotecan in our current study. Nonetheless, the rates of grade 3+ toxicity appear acceptable for this feasibility study. It is important to note, however, that more recent literature has explored modified and genotyped dosing strategies of FOLFIRINOX, including omission of 5-FU bolus and dose reduction to irinotecan 150 mg/m² irinotecan as standard of care in pancreas cancer, and therefore, these modified regimens of FOLFIRINOX would likely further improve tolerability of this regimen with likely no detriment in benefit (14, 19). These studies support the safety and early clinical activity of FOLFIRINOX in gastroesophageal adenocarcinoma.

Moreover, in our small series, the pCR rates (28% in the ITT group and 35% in the resected group) compare favorably with those achieved with other currently accepted neoadjuvant strategies. In the FLOT 4 study, which evaluated the efficacy of perioperative 5-FU, oxaliplatin and docetaxel, the rate of pCR reported after four cycles of neoadjuvant FLOT was 16% (4). More recently, perioperative FOLFIRINOX yielded a pCR rate of 11% (19). As shown in the POET study, the addition of neoadjuvant chemoradiation compared with neoadjuvant chemotherapy alone improves rates of pCR (34). Therefore, it is perhaps not surprising that among studies which include neoadjuvant radiation, the pCR rate was 23% (9) in the CROSS trial and 26% for RTOG 9904 (35), respectively. In evaluating pCR rates among neoadjuvant gastric strategies, it is important to consider the incidence of diffuse type and signet ring subtype within these studies, as these histologies have generally poor response to treatment. Numerous studies have suggested a correlation between pCR and histologic treatment response and long-term disease outcomes (35, 36). In addition to pCR, nodal clearance has also been suggested to be prognostic in the setting of perioperative therapy (37). In our current study, although 80% of patients at baseline had clinically involved LNs, only 30% were found with pathologically involved LNs, suggesting the impact of treatment on nodal clearance. In addition, in our study, the ITT R0 resection rate was 80%, which compares favorably with ITT R0 resection rates seen with perioperative FLOT or ECX (R0 rates of 85% vs. 74%, respectively; refs. 4, 5). In our study, among patients who underwent surgical resection, all had a R0 resection (100%), compared with 89.6% (301/336)and 88.9% (279/314) for FLOT and ECX, respectively (4).”

With a median follow-up of 23.1 months, among the entire cohort, we reported a 2-year DM of 37%, 2-year PFS of 55%, and 2-year OS rate of 72%. These results compare favorably with the 2-year outcomes reported with perioperative FLOT in which 2-year disease-free survival (DFS) (extrapolated from DFS KM curve) was approximately 56% and the 2-year OS was 68% (4). Although this was a small study, we feel that these results are promising and this regimen merits further evaluation in larger cohort of patients.

Of note, surveillance endoscopy in post treatment follow-up was not routinely performed, and only performed as clinically indicated. Therefore, it is possible that occult asymptomatic local recurrence could be missed as there can be latency between recurrence and development of symptoms. Longer term follow-up of the patients enrolled in this current study will be necessary to determine the impact of dose intensification of FOLFIRINOX on long-term distant disease control.

In our current study, we included patients with limited stage nodal metastases which could be encompassed within a radiotherapy treatment field. At our institution, for patients with limited nodal metastatic disease that could be encompassed within an radiotherapy treatment field (and therefore treated with definitive intent chemoradiation dosing), we have often approached these patients with curative intent. There is emerging data regarding the potential role for aggressive local for limited stage metastatic disease. Recently, the AIO-FLOT3 study evaluated patients with limited stage metastatic disease (n = 60), including only retroperitoneal LN involvement [27 patients (45%)], liver involvement [11 (18.3%)], lung involvement [10 (16.7%)], localized peritoneal involvement [4 (6.7%)], or other [8 (13.3%)] incurable sites. They reported a median OS of 22.9 months [95% confidence interval (CI), 16.5–upper level not achieved] in this cohort of limited stage metastatic disease with a response rate of 60% (complete, 10%; partial, 50%). The median OS was 31.3 months (95% CI, 18.9–upper level not achieved) for patients who proceeded to surgery and 15.9 months (95% CI, 7.1–22.9) for the other patients (38). Based on these data, the RENAISSANCE (AIO-FLOT5) study is currently underway evaluating the benefit of surgery on survival and quality of life among patients with nodal metastases or limited volume metastatic disease (39). Future studies will hopefully further clarify the benefit of aggressive local strategies of ablative radiotherapy and surgical resection for limited stage metastatic disease.

Although our current study had limited events, ctDNA appears as a promising predictor of disease recurrence. Despite the small number of patients and events, we found that detectable ctDNA after chemoradiation (P = 0.004) and postoperatively (P = 0.045) were associated with disease recurrence. Our ctDNA analysis was not based on NGS but on ddPCR (which is a single candidate approach), which limits the number of mutations that can be tested on a sample due to limited material. In our study, all but one of the patients were tested by ddPCR for one mutation only, and with one case being tested for two mutations. The mutation that was chosen from the SNaPshot analysis on tissue was the one with the highest frequency, therefore allowing a higher chance to detect it in the ctDNA as well. While our sensitivity may be enhanced by approaches using multiple mutations, sensitivity is also dependent on stage, differences in tumor biology, and amount of plasma analyzed (38). Nonetheless, our data show post-chemoradiation and postoperative ctDNA appears to be a promising biomarker to predict disease recurrence. Analysis of ctDNA after completion of definitive therapy has been shown to be promising in detection of minimal residual disease in several disease sites to date (40–46). Recent retrospective data from Stanford has emerged that suggest that post-chemoradiation ctDNA may identify patients with localized esophageal cancers at increased risk of recurrence and death and may potentially help in the determination of which patients are most likely to benefit from surgery (44). Our study, however, represents to our knowledge the first prospective series to demonstrate the prognostic value of ctDNA in patients with locally advanced esophagogastric cancer treated with total neoadjuvant therapy.

Limited molecular data are available to assess molecular factors that contribute to tumor evolution and subclonal selection during chemoradiation for gastric and gastroesophageal adenocarcinoma tumors (30). While analysis of large tumor bank sequencing data for untreated gastroesophageal adenocarcinoma and gastric tumors has revealed the presence of diverse tumor subtypes within a given anatomic location (47), it is unclear how these subtypes or their molecular hallmarks might contribute to treatment resistance or to the evolution of tumors over the course of treatment. Our exploratory molecular analysis of pre- and post-treatment tumor samples indicates numerous putative or potential models of tumor evolution and suggests heterogeneous responses of subclonal populations to neoadjuvant chemoradiation. These models are suggested by our data, but limitations of our technique prevent conclusive establishment of tumor evolution patterns in this study. Our exploratory analysis is limited by reduced sensitivity due to variable sample purity and the acquisition of only a single analyzed sample for each timepoint, which could lead to undersampling of the full complement of subclones and an underappreciation of their diverse molecular features and complex evolutionary dynamics in any individual tumor (48). Other sequencing efforts have revealed spatial variability of genomic and epigenomic subclonal architecture within gastric tumors (49–52), colorectal cancers (53, 54), and other tumors (55–58), so future studies should be undertaken with attention to more comprehensive sampling of spatially distinct tumor regions. Ultimately, our data provide hypothesis generating support for studies incorporating multiple spatially distinct biopsies in future sequencing, single-cell, and transcriptomic analyses from pre- and post-treatment samples to characterize patterns of treatment adaptation and resistance to inform individualized therapeutic approaches.

In conclusion, in this prospective single-institution pilot study, neoadjuvant FOLFIRINOX followed by chemoradiation is feasible and safe, and encouraging preliminary antitumor activity is highlighted by the high pCR rate and nodal clearance at surgery. A follow-up phase II study of neoadjuvant NAPOX (5-FU, oxaliplatin, and liposomal irinotecan) followed by chemoradiation with paclitaxel and carboplatin in locally advanced esophagogastric cancer (NCT04656041) is currently enrolling at our institution. In combination with the ongoing CRITICS-2, TOPGEAR, ESOPEC, and NeoAEGIS studies, our studies hopes to further inform the optimal treatment strategies for locally advanced gastroesophageal adenocarcinoma cancers Finally, although our current study is limited by events, post-chemoradiation and postoperative ctDNA appears to be a promising biomarker to predict disease recurrence. Future studies investigating the role of ctDNA and evaluation of tumor evolution in the management of gastric and gastroesophageal adenocarcinoma cancers are warranted to corroborate our findings.

J.Y. Wo reports grants from NIH Proton Beam NCI/Federal Share Program grant and Cancer Clinical Investigator Team Leadership Award awarded by the NCI during the conduct of the study. M. Mino-Kenudson reports personal fees from H3 Biomedicine, AstraZeneca, and Elsevier and grants from Novartis outside the submitted work. S.J. Klempner reports personal fees from Eli Lilly, Merck, BMS, Astellas, Daiichi-Sankyo, Pieris, and Natera and other support from Turning Point Therapeutics outside the submitted work. A.R. Parikh reports other support from C2i, Puretech, PMV, Plexxicon, Takeda, BMS, and Novartis; personal fees from Eli Lilly, Pfizer, Checkmate Pharma, Roche, and Natera during the conduct of the study. E. Roeland reports personal fees from Helsinn Therapeutics, Napo Pharmaceuticals, and Asahi Kasei Pharmaceuticals outside the submitted work. D.P. Ryan reports other support from Acworth Pharma and Exact Sciences; personal fees and other support from MPM; and personal fees from Boehringer Ingelheim and McGraw Hill during the conduct of the study; and other support from Acworth Pharma, Exact Sciences and McGraw Hill and personal fees and other support from MPM and Boehringer Ingelheim outside the submitted work. R.B. Corcoran reports personal fees from Abbvie, Asana Biosciences, Astex Pharmaceuticals, AstraZeneca, Elicio, Fog Pharma, Guardant Health, Ipsen, Mirati Therapeutics, Qiagen, Roivant, Taiho, and Zikani Therapeutics; other support from Avidity Biosciences and Erasca; personal fees and other support from C4 Therapeutics, Kinnate Biopharma, nRichDx, Remix Therapeutics, and Revolution Medicines, and grants from Lilly and Novartis outside the submitted work. E. Van Seventer reports other support from Blueprint Medicines outside the submitted work. M. Lanuti reports personal fees from AstraZeneca, Bristol Myers Squibb, and Precisca outside the submitted work. R.S. Heist reports personal fees from Novartis, Daichii Sankyo, EMD Serono, Apollomics, and Tarveda; grants from Novartis, Daichii Sankyo, Mirati, Turning Point, Abbvie, Agios, Corvus, Exelixis, and Lilly outside the submitted work. T.S. Hong reports personal fees from Synthetic Biologics, Novocure, Merck, and PanTher Therapeutics; other support from Taiho, AstraZeneca, BMS, Tesaro, Ipsen, and Puma outside the submitted work. No disclosures were reported by the other authors.

J.Y. Wo: Data curation, formal analysis, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J.W. Clark: Conceptualization, resources, methodology, writing–review and editing. C.E. Eyler: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing–review and editing. M. Mino-Kenudson: Data curation, validation, methodology, writing–original draft. S.J. Klempner: Investigation, writing–original draft, writing–review and editing. J.N. Allen: Writing–original draft, writing–review and editing. F.K. Keane: Resources, data curation, writing–original draft, writing–review and editing. A.R. Parikh: Investigation, writing–original draft, project administration, writing–review and editing. E. Roeland: Investigation, writing–original draft, project administration, writing–review and editing. L.C. Drapek: Data curation, validation, investigation, writing–original draft, writing–review and editing. D.P. Ryan: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing–review and editing. R.B. Corcoran: Resources, data curation, formal analysis, supervision, writing–original draft, writing–review and editing. E. Van Seventer: Resources, data curation, writing–original draft, writing–review and editing. I.J. Fetter: Resources, data curation, investigation. H.A. Shahzade: Data curation, validation, investigation. M.J. Khandekar: Investigation, writing–original draft, writing–review and editing. M. Lanuti: Investigation, writing–original draft, writing–review and editing. C.R. Morse: Investigation, writing–original draft, writing–review and editing. R.S. Heist: Investigation, writing–original draft, writing–review and editing. C.A. Ulysse: Formal analysis, methodology, writing–original draft, writing–review and editing. B. Christopher: Data curation, project administration. C. Baglini: Data curation, project administration. B.Y. Yeap: Conceptualization, resources, data curation, formal analysis, methodology, writing–original draft, writing–review and editing. J.T. Mullen: Conceptualization, resources, formal analysis, methodology, writing–original draft, writing–review and editing. T.S. Hong: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing–review and editing.

This study was funded by NIH Proton Beam NCI/Federal Share Program grant C06 CA059267, and in part by the Cancer Clinical Investigator Team Leadership Award awarded by the NCI though a supplement to P30CA006516 (T.S. Hong)

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.