Background
Esophageal cancer (EC) is the seventh most common malignant tumor and the sixth leading cause of cancer-related mortality worldwide [
1]. In China, approximately 90% of EC cases are squamous cell carcinoma [
2,
3]. Esophagectomy plays a primary role in treating locally advanced esophageal squamous cell carcinoma (ESCC). However, 45% of the patients with surgery alone experience local recurrence or distant metastasis within 5 years after surgery [
4‐
6].
Neoadjuvant therapy could increase the rate of R0 resection and improve survival compared with surgery alone [
6]. Based on the results of the NEOCRTEC5010 trial [
6], chemoradiotherapy followed by surgery has been recommended as the standard treatment for locally advanced ESCC. However, radiotherapy can increase perioperative complications and mortality. Additionally, despite the high R0 resection rate with chemoradiotherapy, about 15% of ESCC patients still suffer from local recurrence within 5 years after surgery, and that rate of distant metastasis reaches 30% [
6,
7]. Therefore, novel strategies are needed to achieve better safety profiles and optimal survival outcomes.
As a newcomer to cancer treatment, programmed cell death 1 (PD-1) blockade-based immunotherapy exploits a strategy based on immune evasion mechanisms to restore antitumor immunity. Camrelizumab is a humanized high-affinity IgG4-kappa anti-PD-1 monoclonal antibody that has demonstrated efficacy and safety in patients with advanced ESCC [
8,
9]. The randomized phase III ESCORT-1
st study reported that the addition of camrelizumab to chemotherapy improved overall survival (OS) and progression-free survival (PFS) compared with chemotherapy alone, and it has been approved to treat unresectable advanced ESCC with camrelizumab plus chemotherapy in China [
10]. In addition, neoadjuvant use of PD-1 blockade in combination with chemotherapy has also shown favorable antitumor efficacy in several malignancies, including lung [
11,
12] and colorectal cancers [
13,
14]. However, the combination of PD-1 blockade with chemotherapy in locally advanced ESCC has not been well determined.
To date, a few clinical trials have reported that neoadjuvant immunochemotherapy of PD-(L)1 blockade induced a favorable pathological response and tolerant toxicity in patients with locally advanced ESCC [
15,
16]. However, all these studies were small cohort studies with only two cycles. Theoretically, immunochemotherapy has a huge potential to induce long-term tumor regression, eradicate micrometastases, and even cure locally advanced ESCC [
17‐
19]. According to previous data of locally advanced ESCC who received two cycles of PD-1 blockade plus chemotherapy, at a median follow-up of 13 months, recurrence still occurred in approximately 25% of patients, indicating a short course of treatment may not be sufficient [
16]. Indeed, a retrospective study based on real-world data reported that patients with locally advanced ESCC received varying cycles of neoadjuvant immunochemotherapy, with the majority receiving 2–4 cycles. However, the relationship between treatment cycles and pathological responses was not investigated [
20]. Meanwhile, a real-world retrospective study in lung cancer has shown that three and four cycles of neoadjuvant immunochemotherapy were prone to higher major pathological response (MPR) rates than two cycles [
17]. Given the encouraging efficacy and acceptable safety of PD-(L)1 blockade plus chemotherapy in solid tumors, intensive treatment deserves to be explored for optimal clinical outcomes.
In our previous retrospective study [
21], intensive cycles of camrelizumab plus chemotherapy before surgery exhibited promising efficacy without increasing complications in locally advanced ESCC. Therefore, we further performed this phase II trial to evaluate the efficacy and safety of intensive treatment in locally advanced ESCC. Computerized tomography (CT) and safety assessment were conducted before the initiation of treatment and after the second and third courses of neoadjuvant immunochemotherapy to compare the efficacy and safety of two and three treatment cycles. In addition, little is known about biomarkers predicting the efficacy of neoadjuvant immunochemotherapy, which have also been explored in this study.
Methods
Participants
In this single-center, single-arm, phase II trial, camrelizumab was combined with chemotherapy followed by surgery for locally advanced ESCC. Inclusion criteria were (1) stage II or III locally advanced resectable ESCC diagnosed before enrollment (2) no distant organ metastases or cervical lymph node metastases prior to enrollment (3) no secondary primary tumors (4) an Eastern Cooperative Oncology Group (ECOG) performance status score 0 or 1 (5) no prior exposure to anticancer therapy, including chemotherapy, radiotherapy, targeted therapy, and immunotherapy.
Procedure
Participants were administered three cycles of chemotherapy and PD-1 blockade. For each cycle of treatment, patients were intravenously administered a flat dose of camrelizumab (200 mg) along with a single dose of nab-paclitaxel (260 mg/m
2) on day 1, and capecitabine was orally administered twice daily (1250 mg/m
2) on days 1 through 14. The regimen was repeated every 3 weeks (Additional file
1: Fig. S1). A prophylactic dose of granulocyte colony-stimulating factor (G-CSF) was administered on day 4 of each cycle. The following tests were performed at baseline, two and three times after the neoadjuvant treatment cycles: contrasted-enhanced thoracic/abdominal CT, endoscopic ultrasonography (EUS), and cervical/subclavicular ultrasonography. Radiographic responses of primary tumors were evaluated using CT scan images acquired before and after two and three cycles of neoadjuvant treatment according to Response Evaluation Criteria in Solid Tumors version (RECIST) 1.1 [
22]. All imaging data were reviewed by two independent radiologists. Treatment-related adverse events (TRAEs) were reported according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE), version 5.0, at each visit [
23].
In approximately four to six weeks after the last course of neoadjuvant therapy, a thoracoscopy esophagectomy was performed with cervical esophagogastric anastomosis and total dissection of two-field lymph nodes (LNs). The removal of lymph nodes included recurrent laryngeal nerve nodes, subcarinal nodes, paraesophageal nodes, pulmonary ligament nodes, cardia nodes, left gastric artery nodes, and lesser curvature nodes. Surgical sections were stained with hematoxylin and eosin (H&E), and pathological regression was assessed by two independent pathologists. Complete pathological response (pCR) was defined as the absence of residual invasion disease. Tumors with ≤ 10% residual viable tumor cells were considered as obtaining an MPR.
After surgery, follow-up was conducted every 3 months in the first year, every 6 months for the second and third years, and every 12 months thereafter. Overall survival (OS) was defined as the time between the surgery and the end of follow-up or death. Disease-free survival (DFS) was calculated from the surgery date to the end of follow-up or the date of the first recurrence.
Outcome
The primary endpoint of the study was pCR. The secondary endpoints included safety, feasibility, MPR, radiologic response, DFS, and OS.
Exploratory analysis
Pretreatment tumor biopsy was obtained using EUS for biomarker analysis, including programmed cell death-ligand 1 (PD-L1) expression, tumor mutational burden (TMB), and tumor immune microenvironment (TIME). PD-L1 expression was assessed using the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies). The combined positive score (CPS) was used to define PD-L1 expression, which was determined by dividing the number of PD-L1-positive tumor and immune cells by the total number of viable tumor cells and multiplying by 100. Next-generation sequencing (NGS) was performed using whole-exome sequencing or a 733-gene panel (3D Medicines Inc.). As defined, the TMB was the number of somatic single nucleotide variations (SNVs) and insertions/deletions (indels) per megabase of coding genome sequenced. Synonymous and non-synonymous mutations, stop gains/losses, and splicing variants were all considered SNVs. Indels included both frameshift and non-frameshift insertions and deletions. Non-coding alterations were excluded from the calculation of TMB. TIME was evaluated using multiplex immunofluorescence (mIF) staining. The quantities of CD8+ T cells, tumor-associated macrophages (TAMs), and natural killer (NK) cells were expressed as the number of stained cells per square millimeter. Posttreatment tissue was also collected and subjected to mIF to analyze the change in the TIME after neoadjuvant immunotherapy. Besides, the posttreatment tissues were also submitted to H&E staining and immunostaining for CD3 and CD20 to analyze the tertiary lymphoid structures (TLSs).
Statistical analyses
This study applied superiority designs with the primary endpoint of pCR. According to previous studies, the pCR rate of chemotherapy is hypothesized to be 15% [
24,
25]. With the consideration of a dropout rate of 10%, a total of 47 patients would need to be enrolled to provide 80% power to detect a pCR of 34% at a one-sided 5% alpha level. Continuous variables were compared using the Mann–Whitney U test, and categorical variables were compared using the chi-square or Fisher exact test, as appropriate. All reported
P values were two-tailed. A
P value of < 0.05 was considered statistically significant. Survival curves were estimated using the Kaplan–Meier method. All analyses and graph generation were performed using R 3.6.0.
Discussion
In this phase II trial conducted in 47 patients with locally advanced ESCC, the pCR and MPR were 33.3% and 64.3%, respectively, and the ORR was 80.0%. Forty-two patients received surgery, and R0 resection was achieved in 100% of patients having undergone surgery. Three treatment cycles elicited a significantly higher rate of T down-staging than two (84.4% vs. 62.2%) without a significant increase in TRAEs. The most common TRAEs were grade 1–2, and no surgical delay was reported. With a median follow-up of 24.3 months, the 1-year DFS and OS rates were both 97.6%, and the 2-year DFS and OS rates were 92.3% and 97.6%, respectively. The density of CD56dim NK cells in the pretreatment tissues was significantly higher in the pCR group than in the non-pCR group. While the density of TLSs in the posttreatment tissues was numerically higher in the pCR group, and this difference became statistically significant when comparing patients with MPR to those with non-MPR. No difference was found in PD-L1 expression and TMB levels between pretreatment specimens of the pCR and non-pCR patients.
Neoadjuvant therapy is recommended in many cancers to achieve tumor downstaging and improve the curative rate. However, different cycles of neoadjuvant treatment could influence the prognosis and the quality of perioperative life. In a randomized phase II study [
28], three courses of preoperative chemotherapy led to a better response without increasing TRAEs or morbidity than two courses in ESCC. For neoadjuvant immunochemotherapy, all the available clinical trials have focused on the effects of two-cycle regimens, which could be efficiently limited. In our previous pilot study [
21], three cycles of neoadjuvant immunochemotherapy was safe and feasible, which was further confirmed in this phase II trial.
The toxicity of intensive cycles of camrelizumab plus chemotherapy was tolerated. Most of the TRAEs were grade 1–2, which was similar to previous data of two treatment cycles [
15,
16]. RCCEP was found in 28 (59.6%) patients. The incidence was higher than those of the two treatment cycles (26.1–39.1%) [
15,
16]. This difference may have resulted from the additional course of camrelizumab. In addition, more than half of patients experienced leukopenia after receiving neoadjuvant immunochemotherapy or chemoradiotherapy, according to previous reports [
6,
16]. Severe leukopenia can even lead to dose reduction or termination of treatment. In our study, there was only a low frequency (6.4%) of leukopenia. The difference could be attributed to the following reasons. First, we prophylactically used G-CSF after each course of chemotherapy treatment. Second, platinum was replaced with capecitabine in our regimen. Capecitabine is an oral drug and can be converted to fluorouracil [
29]. The combination of capecitabine with paclitaxel exhibited similar efficacy but lower toxicity compared with platinum-based regimens in breast cancer and head and neck squamous cell carcinoma [
30‐
32]. The toxicity of this drug is relatively low, which might render it a suitable candidate for combination with PD-1 blockade. Overall, the toxicity of our neoadjuvant regimen was manageable and worthy of promotion.
For surgery completion, R0 resection was achieved in 100% of patients who underwent surgery, which was consistent with that of other two-cycle regimens (96.3–100%) [
16,
33]. The volumes of lymph node dissection far exceeded those of other two-cycle camrelizumab treatment [
16]. It seems that one additional course of immunotherapy would not increase the difficulty in conducting surgery and lymph node dissection. With regard to complications, anastomotic leakage was the most frequent complication, with an incidence of 19.0%, which was in the normal range compared with surgery alone (15% to 20%) [
34]. Moreover, the time of operation duration and patient hospital stays were not prolonged. No perioperative deaths were reported in our cohort. These results suggested that the intensive-cycle regimen was feasible.
In this study, the pCR rate was 33.3%, similar to the results from other immunochemotherapy trials of ESCC. Taking an intensive-cycle regimen does not seem to impact the pCR (data from two-cycle immunochemotherapy studies: 25–35.3%) [
16,
35‐
37]. Whereas, CT assessment conducted at treatment milestones (before and after the second and third course of neoadjuvant therapy) indicated that three treatment cycles elicited a significantly higher rate of T down-staging than two (84.4% vs. 62.2%,
P = 0.03), without increasing TRAEs, suggesting the feasibility and safety of three cycles of immunochemotherapy to increase tumor regression. These results were consistent with data from locally advanced lung cancers [
17,
38].
Furthermore, our study found that 2.4% (1/42) of patients developed local recurrence and 4.8% (2/42) experienced distant metastasis at a median follow-up time of 24.3 months after surgery, which were numerically lower than the respective rates of 20% (4/20) and 10% (2/20) observed in patients who received two cycles of neoadjuvant PD-1 blockade plus chemotherapy at a median follow-up time of 13.5 months [
16]. Both the 1-year OS (97.6% vs. ~ 90.0%) and DFS (97.6% vs. ~ 80.0%) were numerically higher than those with two-cycle immunochemotherapy regimens [
16,
36]. The potential explanation for these data was that except for the advantage of increasing tumor shrinkage, intensive cycles of immunochemotherapy might exert longer-term antitumor activity, thereby inducing longer-term tumor regression and eradicating micrometastases. Follow-up is ongoing, and long-term survival data will be released in the future.
The pathological response was significantly predictive of prognosis [
39]. It is essential to explore biomarkers to identify patients who might benefit from the treatment. PD-L1 and TMB were the most commonly investigated biomarkers. Our study found that the level of PD-L1 expression and TMB at baseline had poor correlations with the pathological response, which was consistent with previous studies that PD-L1 and TMB failed to precisely predict the efficacy of neoadjuvant immunotherapy [
40,
41]. Instead, CD56
dim cells were found at a higher density in the pretreatment biopsy of responders. This observation was consistent with our previous finding [
21]. CD56 cells are the major subtype of NK cells and the primary force of innate immunity for anti-tumor response [
42]. Moreover, we observed a decrease in the density of NK
dim cells in the stroma after immunotherapy in responders but not in non-responders. The decrease in the infiltration of CD56 cells in the stroma might have been attributed to the fact that PD-1 blockade could induce the mobilization of more abundant NK cells to infiltrate from the stroma to the parenchyma. Previous work in melanoma supported that immune cells infiltrated from the tumor edge and gradually infiltrated to the core of the tumor upon immunotherapy treatment [
43]. Furthermore, TLSs were found to be more abundant in surgical tissues from pCR or MPR patients than those from non-pCR or non-MPR patients, which was consistent with previous reports linking TLSs to a favorable prognosis following immunotherapy [
26,
27].
Molecular genetic analyses demonstrated multiple genetic abnormalities in ESCC. Our study found some specific driver mutations, including
CCND1,
FGF19, and
FGF4. These genes were located in 11q13, which has been considered the most frequently amplified locus in ESCC and is related to the development of ESCC [
44]. However, all the driver mutations failed to predict the response to immunotherapy. This might have resulted from the complex genomic context in locally advanced ESCC. It would be difficult to predict prognosis with a single gene.
To summarize, intensive cycles of neoadjuvant chemotherapy combined with camrelizumab demonstrated favorable efficacy and acceptable toxicity, particularly an encouraging 1-year DFS and OS. The abundance of CD56dim NK cells in the pretreatment tumor tissue might be a potential biomarker to predict the efficacy of immunotherapy in locally advanced ESCC. The follow-up of this study is still ongoing, and the long-term survival data will be released in the future. Due to the limited sample size and the single-arm manner of the study, randomized controlled trials with larger sample sizes are needed to confirm our findings.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.