The safety of current treatment options for advanced esophageal cancer after first-line chemotherapy

ABSTRACT

The prognosis of advanced esophageal cancer (EC) remains poor, and few effective agents are available. For advanced EC patients, a combination of platinum and fluoropyrimidine is recognized as the standard first-line treatment. After first-line treatment, taxane or irinotecan has been used. Based on the KEYNOTE-181 and the ATTRACTION-3 trials, immune checkpoint inhibitors (ICIs) such as pembrolizumab and nivolumab appear to prolong survival, compared with cytotoxic agents, as second-line treatments for advanced EC patients. In addition, ICIs have different safety profiles than conventional cytotoxic agents. Herein, we discuss the differences in the safety profiles of cytotoxic agents and ICIs for the treatment of advanced EC patients after first-line chemotherapy.ICIs as a second-line treatment are tolerable in advanced EC patients. Although infrequent, ICIs can cause immune-related adverse events that are sometimes fatal. Therefore, regular monitoring of physical and laboratory examinations is needed during and after the administration of ICIs. As the major toxicities of taxane are neutropenia and neuropathy, while those of irinotecan are neutropenia and diarrhea, appropriate supportive care or dose modification may be needed for individual patients. ICI-containing treatments have been developed not only as second-line treatments, but also as first-line treatments or for use in perioperative settings; thus, particular attention with regard to immune-related toxicities is needed.

KEYWORDS:

• Esophageal cancer
• immune-related adverse event
• taxane
• irinotecan
• nivolumab
• pembrolizumab
• safety

1. Introduction

Esophageal cancer (EC) is the seventh most common cancer (570,000 new cases/year) and the sixth highest cause of cancer-related mortalities (509,000 deaths/year) worldwide[1]. The major histological subtypes of EC are squamous cell carcinoma and adenocarcinoma. Esophageal squamous cell carcinoma (ESCC) comprises over 90% of all ECs in the world, and ESCC is a more common histological subtype than esophageal adenocarcinoma (EAC) in Eastern Asia and Eastern Africa. On the other hand, in Western countries, EAC is the most frequent histological subtype. The major risk factors differ between ESCC and EAC. Alcohol consumption and smoking are the most important risk factors for ESCC, while the most important risk factors associated with EAC are Barret’s esophagus, obesity and chronic gastroesophageal reflux disease. EC patients are often asymptomatic, but oral intake can be difficult in patients with advanced disease because of stenosis. The rate of the five-year survival of EC patients is reported to be 10%-40% [2–6].

Effective treatment options for metastatic or recurrent EC are limited. For metastatic or recurrent EC patients, systemic chemotherapy is the standard treatment [7–9]. Recently, clinical trials have shown immune checkpoint inhibitors (ICIs) to provide a survival benefit when administered as second-line treatments for EC [5,10]. In clinical practice, therefore, oncologists must be aware of the immune-related adverse events (irAEs) that differ from conventional adverse events, such as cytopenia or nausea, caused by cytotoxic agents.

In this review, we will discuss the differences in the safety profiles of conventional cytotoxic agents and ICIs for the treatment of metastatic or recurrent EC after first-line chemotherapy.

2. Standard treatments for metastatic or recurrent EC

Systemic chemotherapy is the standard treatment for metastatic or recurrent EC, and platinum plus fluoropyrimidine doublet chemotherapy is recognized as a first-line treatment. The response rate for cisplatin plus 5-FU chemotherapy as a first-line treatment is reportedly about 30%, and the median overall survival (OS) is about 6.6–10.4 months [6,11–13]. After the failure of first-line doublet chemotherapy, nivolumab is the standard second-line chemotherapy for ESCC patients, while pembrolizumab is only used in patients with a combined positive score (CPS) ≥10 [5,10].

The ATTRACTION-3 phase III trial was conducted to compare nivolumab with taxane (paclitaxel or docetaxel) after a refractory or intolerant response to one previous course of fluoropyrimidine and platinum-based therapy for metastatic or recurrent ESCC. Eligible patients were age 20 years or older and performance status score of 0 or 1 whose major current or previously resected lesion was was pathologically confirmed as squamous or adenosquamous cell carcinoma. Patients were refractory or intolerant to fluoropyrimidine-based and platinum-based chemotherapy who had previously received one treatment regimen. This trial showed the superiority in terms of OS in the nivolumab group, compared with the taxane group (median OS: 10.9 months vs. 8.4 months, HR [95% CI]: 0.77 [0.62–0.96], p = 0.019).

The KEYNOTE-181 phase III trial was conducted to compare pembrolizumab with chemotherapy (taxane or irinotecan) as a second-line treatment for metastatic or recurrent esophageal and EGJ cancer, including 64% of ESCC patients. Eligible patients were ≧ 18 years of age and performance status of 0 or 1 with histologically confirmed squamous cell carcinoma or adenocarcinoma of the esophagus. Selection criteria included metastatic or locally advanced, unresectable disease, that progressed after one standard chemotherapy. Median OS was 9.3 months (95% CI, 6.6 to 12.5 months) in pembrolizumab group and 6.7 months (95% CI, 5.1 to 8.2 months) in chemotherapy group, respectively. Pembrolizumab did not result in a superior OS, compared with cytotoxic agents, for all the prespecified primary endpoints; however, pembrolizumab has only been approved for use in ESCC patients with a PD-L1 CPS≥10.

Based on the results of these two trials, ICIs became a second-line standard treatment for ESCC patients. However, for EAC patients, paclitaxel monotherapy or combined treatment with ramucirumab are used in clinical practice according to the results of the two phase III trials:　RAINBOW and KEYNOTE-061 trials [14,15]. In a meta-analysis report, these cytotoxic agents were shown to have clinical benefits, compared with best-supportive care[16].

3. Cytotoxic agents

3.1. Paclitaxel

Paclitaxel is a cytotoxic antitumor drug that binds to microtubules, which are important during the mitotic phase of the cell cycle where paclitaxel shows antitumor activity. Paclitaxel has become an important agent in the treatment of several cancers such as ovarian carcinoma, breast cancer, lung cancer and EC. The administration dose of paclitaxel varies according to indication and body weight and is reduced in cancer patients with liver dysfunction. The characteristic side effects of paclitaxel are reportedly hypersensitivity, peripheral neuropathy, pneumonitis, bradycardia, myalgias, arthralgias, alopecia, myelosuppression, and a transient elevation in serum transaminases[17].

A phase II clinical trial of paclitaxel monotherapy showed promising anti-tumor activity for metastatic or recurrent EC patients who were previously treated with platinum-based chemotherapy[18]. Eligible patients had been histologically diagnosed as having ESCC, EAC, or esophageal adenosquamous carcinoma and had developed either metastatic or recurrent disease after platinum-based chemotherapy. Patients received paclitaxel (100 mg/m²) once a week on days 1, 8, 15, 22, 29, and 36 intravenously over 1 h every 7 weeks. Fifty-six patients were enrolled, and 3 patients were withdrawn prior to receiving paclitaxel because of their hemoglobin levels. Fifty-two of the 53 patients had ESCC. As the primary endpoint, the overall response rate was 44.2% (23/52) in all evaluable patients. The median progression-free survival (PFS) was 3.9 months (95% CI: 2.8–4.6 months), and the median OS was 10.4 months (95% CI: 7.8–14.2 months,). In the ATTRACTION-3 phase III trial, 144 patients in the control arm received paclitaxel (100 mg/m²) at the same interval as the above-mentioned phase II trial. A lower dose and schedule (80 or 100 mg/m², once a week on days 1, 8, and 15, every 4 weeks) was used for the KEYNOTE-181 trial. The efficacies of taxanes were comparable or slightly lower in the phase III study than in the phase II study.

In the safety profile for the phase II trial, the proportion of grade 3/4 neutropenia was 52.8%. On the other hand, common non-hematologic toxicities of all grades were alopecia (83.0%), sensory neuropathy (81.1%), and fatigue (71.1%). Grade 3 or 4 non-hematologic toxicities were pneumonia (7.5%), sensory neuropathy (5.7%), febrile neutropenia (3.8%) and intestinal lung disease (3.8%), respectively. One patient showed a grade 4 hypersensitivity reaction. Additionally, 18 patients (34.0%) discontinued paclitaxel monotherapy because of adverse events such as myelosuppression and sensory neuropathy, 23 patients (43.4%) required a dose reduction of paclitaxel because of adverse events such as sensory neuropathy, and the administration of paclitaxel was skipped or delayed because of neutropenia in 28 patients (52.8%). Finally, no treatment-related deaths occurred in this trial.

In the ATTRACTION-3 phase III trial, no adverse events were described for either paclitaxel or docetaxel, but adverse events were described for taxanes. The common toxicities (all grades/grade ≥3) of the taxanes were alopecia (29.1%/0.3%), neutropenia (11.5%/7.1%), peripheral sensory neuropathy (16.9%/0.3%), and febrile neutropenia (-/10.6%)[5]. In the KEYNOTE-181 trial, common toxicities (all grades/grade ≥3) of chemotherapy were alopecia (47.1%/-), neutropenia (36.5%/28.4%), and peripheral sensory neuropathy (22.6%/0.5%). Importantly, these incidences do not represent the data for each chemotherapy regimen, but for all patients who received paclitaxel, docetaxel, or irinotecan monotherapy[10].

Based on these data, paclitaxel monotherapy might cause frequent neutropenia as a hematological toxicity and sensory neuropathy as non-hematological toxicity. First, serious neutropenia can sometimes cause a fatal infection, such as sepsis or febrile neutropenia, and the white cell and neutrophil counts should always be checked before the administration of paclitaxel in clinical practice, with appropriate adjustments such as dose modification or the discontinuation of paclitaxel as needed. Second, sensory neuropathy can reduce the quality of life of EC patients, potentially leading to the discontinuation of treatment with effective agents. Additionally, some patients with advanced EC who received an oxaliplatin-containing regimen as a first-line chemotherapy might continue to exhibit sensory neuropathy at the time of paclitaxel administration. Therefore, a physical examination for neuropathy is needed, and appropriate adjustments such as dose modification or the discontinuation of paclitaxel may be necessary.

3.2. Docetaxel

Docetaxel is a semisynthetic analogue of paclitaxel. Docetaxel is a cytotoxic drug that produces an antitumor effect by binding to microtubules during the mitotic phase of the cell cycle and that plays an important role in the treatment of several cancers including breast, lung, and esophageal cancer. The characteristic adverse events of docetaxel are neutropenia, hypersensitivity reaction, fluid retention, skin rush, alopecia, peripheral neuropathy, fatigue, and myalgias[17], similar to those of paclitaxel.

A phase II clinical trial of docetaxel monotherapy in Japanese patients with metastatic or recurrent EC has been reported[19]. Eligible patients had been previously treated with or without one regimen of chemotherapy. Docetaxel (70 mg/m²) was infused over 1–2 h every 3 weeks. Among fifty-two EC patients, the overall response rate was 20% (95% CI: 10%-34%). The median OS was 8.1 months (95% CI: 6.6–11.3 months), and the 1-year survival rate was 35% (95% CI: 21%-48%). In the ATTRACTION-3 and KEYNOTE-181 trials, docetaxel (75 mg/m²) every 3 weeks was used for the patients in the control arm.

As for the hematologic toxicity, the frequencies of all-grade and grade 3 or 4 neutropenia were reported to be 100.0% and 87.8%, respectively. On the other hand, common non-hematologic toxicities of all grades were fatigue (89.8%), anorexia (73.5%), diarrhea (40.8%), nausea (36.7%), febrile neutropenia (18.4%) and vomiting (16.3%). Grade 3 or 4 non-hematologic toxicities were febrile neutropenia (18.4%), anorexia (18.4%), fatigue (12.2%), diarrhea (6.1%) and nausea (4.1%). Overall, 18 patients (36.7%) required a dose reduction of docetaxel because of adverse events, and 28 patients (57.1%) received granulocyte-colony stimulating factor because of neutropenia and febrile neutropenia. No treatment-related deaths occurred in this trial.

In the ATTRACTION-3 trial and the KEYNOTE-181 trials, the toxicities of taxanes were previously described. While we cannot evaluate the different safety profiles between paclitaxel and docetaxel based on the results of the ATRRACTION-3 and KEYNOTE-181 trials, no significant differences were seen between the phase II and phase III trials.

A retrospective study comparing paclitaxel with docetaxel for metastatic or recurrent EC in patients who had previously received platinum-based chemotherapy showed that the frequencies of grade 3 or 4 neutropenia (48.8% vs. 21.1%, p = 0.003) and febrile neutropenia (20.9% vs. 5.3%, p = 0.029) were higher in the docetaxel group than in the paclitaxel group. On the other hand, the frequency of sensory neuropathy (60.5% vs. 15.1%, p < 0.001) was higher in the paclitaxel group than in the docetaxel group. Additionally, no significant differences in survival outcomes (median PFS: 3.5 months vs. 2.1 months, p = 0.21, median OS: 7.2 months vs. 6.1 months, p = 0.23) were seen between the two groups[20].

Given these data, docetaxel monotherapy might cause neutropenia as a hematological toxicity relatively frequently. Therefore, the white cell and neutrophil counts should be evaluated before the administration of docetaxel in clinical practice; appropriate adjustments, such as dose modification or discontinuation, might be required, similar to paclitaxel monotherapy.

When selecting taxane monotherapy for metastatic or recurrent EC patients after platinum-based chemotherapy in clinical practice, physicians should focus on the different treatment schedule and different safety profiles, such as the risks of neutropenia, febrile neutropenia, and sensory neuropathy. Advanced EC patients who have stenosis caused by the primary tumor often develop aspiration pneumonia. If both aspiration pneumonia due to stenosis and neutropenia occur at the same time, advanced EC patients might have a fatal course. A retrospective study showed that docetaxel monotherapy caused neutropenia and febrile neutropenia more frequently than paclitaxel monotherapy [20]. Additionally, in terms of the treatment schedule, the precise dosage of paclitaxel can be adjusted according to each patient’s condition. Therefore, based on these drug characteristics, we usually select paclitaxel monotherapy as a treatment for advanced EC patients more frequently than docetaxel monotherapy. However, for advanced EC patients with oxaliplatin-induced sensory neuropathy or who are unable to visit the hospital once a week due to schedule of paclitaxel, we often select docetaxel monotherapy.

3.3. Irinotecan

Irinotecan is a semisynthetic derivative of camptothecin, which inhibits the activity of topoisomerase I [21,22]. Irinotecan has become an important agent in the treatment of several cancers such as lung cancer, cervical cancer, ovarian cancer and gastrointestinal cancer[23]. The characteristic adverse events of irinotecan are myelosuppression, diarrhea, nausea, vomiting and pneumonitis. After administration, irinotecan is metabolized to SN-38 in the liver and is then excreted in bile juice in a process that involves UGT1A1. The function of UGT1A1 is mainly affected by UGT1A1*6 and UGT1A1*28, and UGT1A1 polymorphisms are reportedly associated with neutropenia caused by irinotecan in patients with metastatic colorectal cancer and metastatic gastric cancer [24–26].

Small phase II clinical trials of irinotecan monotherapy have been reported for patients with advanced EC, similar to that for taxane[27]. One phase II trial was conducted to evaluate irinotecan monotherapy in unresectable EC patients. Irinotecan (100 mg/m²/body) was infused over 30 min once weekly for 3 weeks followed by 1 week of rest. Fourteen patients were enrolled, 7 patients (50%) had ESCC, and 7 patients (50%) had EAC. Two patients (15%) achieved a PR, the median PFS was 2 months (95% CI: 1-8 months), the median OS was 5 months (95% CI: 1.5-8.5 months), and the 1-year survival rate was 16%.

In the safety profiles evaluated using WHO criteria, the frequencies of all-grades leukocytopenia and anaemia were 21.4% and 28.6%, respectively. Additionally, Grade 3 or 4 hematologic toxicity was infrequent, and the frequencies for leukocytopenia and anaemia were 0% and 7.1%, respectively. On the other hand, common non-hematologic toxicities of all grades were diarrhoea (57.1%), pain (42.9%), nausea (28.6%), fatigue (28.6%) and anorexia (28.6%). Grade 3 or 4 non-hematologic toxicities consisted of diarrhoea (21.4%), fever (7.1%) and hypercalcaemia (7.1%). Overall, 3 patients (21.4%) required a dose reduction of irinotecan because of adverse events.

These trials did not evaluate UGT1A1 polymorphisms for EC patients, and whether irinotecan-induced neutropenia is associated with the UGT1A1 polymorphism status remains unclear.

4. Immune checkpoint inhibitors

Recently, the efficacy and safety of ICIs have been shown in many clinical trials, and an anti-programmed cell death-1 (PD-1) antibody, which is one type of ICIs, has changed the management of a variety of cancers [28,29].

ICIs have the potential to induce autoimmune effects even in healthy organs; these effects have been termed irAEs. These adverse events often occur in the gastrointestinal tract, endocrine glands, skin, and liver. Less often, the central nervous system, cardiovascular, pulmonary, musculoskeletal, and hematologic systems are involved. The discontinuation of ICIs and/or glucocorticoids and hormone replacement therapy are usually recommended for the management of irAEs according to treatment guidelines [30–32].

4.1. Nivolumab

ATTRACTION-1 trial was an open-label, single-arm, phase II study [33]. Eligible patients had advanced squamous-cell carcinoma, adenosquamous-cell carcinoma, or adenocarcinoma of the esophagus, whose were refractory or intolerant to fluoropyrimidine-based, platinum-based, and taxane-based chemotherapy. Patients were treated with 3 mg/kg nivolumab given intravenously once every 2 weeks in 6-week cycles. Eleven (17%, 95% CI 10–28) of 64 patients showed a objective response. According to the safety profiles of the ATTRACTION-1 trial, the most common treatment-related adverse events (TRAEs) were diarrhea, decreased appetite, rash, and fatigue of any grade. The incidence of grade 1–2 TRAEs and grade 3 or worse were 55% and 17%, respectively. The most common adverse events of grade 3 or worse were decreased appetite (3%), lung infection (3%), increased blood creatine phosphokinase (3%), and dehydration (3%). Seven patients discontinued therapy because of TRAEs, but no deaths related to TRAEs occurred (Table 1).

Table 1. Main outcomes and adverse events in the ATTRACTION-1 trial

	ATTRACTION-1
Efficacy
Number of patients	65
ORR (%)	17
Median PFS (months)	1.5
Median OS (months)	10.8
Treatment-related AEs	n (%)
	Any grade (N = 55)	Grade ≥3 (N = 55)
Diarrhea	9 (14)	0 (0)
Decreased appetite	6 (9)	2 (3)
Constipation	1 (2)	0 (0)
Fatigue	5 (8)	1 (2)
Rash	5 (8)	0 (0)
Elevated blood creatinine phosphokinase level	3 (5)	2 (3)
Dehydration	2 (3)	2 (3)
Pneumonia	1 (2)	0 (0)
Lung infection	4 (6)	2 (3)
Interstitial lung disease	3 (5)	1 (2)
Hyponatremia	1 (2)	1 (2)
Hyperkalemia	0 (0)	0 (0)
Abnormal hepatic function	2 (3)	1 (2)
Bile duct stenosis	0 (0)	0 (0)
Decreased neutrophil count	1 (1)	1 (2)
Events leading to discontinuation	7 (11)
Events leading to death	0 (0)

Abbreviations: ORR, objective response rate; PFS, progression-free survival; OS, overall survival; AE, adverse event

As for the ATTRACTION-3 trial, the most common TRAEs in the nivolumab arm were rash (11.0%), diarrhea (10.5%), decreased appetite (7.7%) and fatigue (7.2%) of any grade. The most common serious TRAEs of any grade in the nivolumab arm were pyrexia (2.4%) and interstitial lung disease (1.9%). One event of grade 4 diabetic ketoacidosis was reported after the data cutoff in the nivolumab group. The incidences of TRAEs leading to discontinuation were almost the same in the nivolumab group (8.6%) and the chemotherapy group (9.1%), but differences in the types of adverse events were observed (Table 2). Five events leading to death occurred: two in the nivolumab group (one each of interstitial lung disease and pneumonitis), and three in the chemotherapy group (one each of pneumonia, spinal cord abscess, and interstitial lung disease).

Table 2. Main outcomes and treatment-related adverse events in the ATTRACTION-3 trial

	ATTRACTION-3
Efficacy	Nivolumab		Chemotherapy
Number of patients	210		209
ORR (%)	19		22
Median PFS (months)	1.7		3.4
Median OS (months)	10.9		8.4
Treatment-related AEs	N = 209 (100)		N = 208 (100)
	Any grade	Grade ≥3	Any grade	Grade ≥3
Anemia	5 (2)	4 (2)	49 (24)	19 (9)
Decreased white blood cell count	2 (1)	1 (<1)	72 (35)	46 (22)
Decreased neutrophil count	3 (1)	1 (<1)	76 (37)	59 (28)
Neutropenia	1 (<1)	0 (0)	40 (19)	29 (14)
Febrile neutropenia	0 (0)	0 (0)	22 (11)	22 (11)
Diarrhea	22 (11)	2 (1)	20 (10)	2 (1)
Nausea	4 (2)	0 (0)	34 (16)	1 (<1)
Stomatitis	5 (2)	1 (<1)	25 (12)	1 (<1)
Decreased appetite	16 (8)	2 (1)	56 (27)	10 (5)
Fatigue	15 (7)	1 (<1)	43 (21)	9 (4)
Alopecia	3 (1)	0 (0)	98 (47)	0 (0)
Arthralgia	3 (1)	0 (0)	21 (10)	1 (<1)
Peripheral sensory neuropathy	1 (<1)	0 (0)	47 (23)	1 (<1)
Rash	23 (11)	1 (<1)	31 (15)	2 (1)
Serious events	33 (16)	20 (10)	47 (23)	41 (20)
Events leading to death	2 (1)	2 (1)	3 (1)	3 (1)
Events leading to discontinuation	18 (9)	8 (4)	19 (9)	13 (6)

Abbreviations: ORR, objective response rate; PFS, progression-free survival; OS, overall survival; AE, adverse event

In addition, an evaluation of safety using patient-reported outcomes and EQ-5D-3 L questionnaires was conducted, and valid responses were obtained from about 85% of the eligible patients. When health-related QOL in the nivolumab arm was compared with QOL in the cytotoxic chemotherapy arm using the EQ-5D-3 L VAS and the utility index, both evaluations showed significantly better results in the nivolumab arm, compared with the cytotoxic chemotherapy arm (EQ-5D-3 L VAS: least squares mean [95% CI], 6.9 [3.0–10.9], p = 0.00069; utility index [95% CI], 0.076 [0.011–0.142], p = 0.002). In addition, when evaluated based on the mean change from the baseline EQ-5D-3 L VAS in both arms, the mean change from the baseline in the nivolumab arm tended to be favorable, compared with that for cytotoxic chemotherapy, until 42 weeks after the start of treatment; the same tendency for the mean change from the baseline EQ-5D-3 L VAS was also observed using the utility index until 42 weeks after the start of treatment. Interestingly, 57% and 55% of the patients in the nivolumab and cytotoxic chemotherapy arms of the ATTRACTION-3 trial received subsequent therapy (radiotherapy: 14% vs. 11%, surgery: 3% vs. 7%, chemotherapy: 53% vs. 47%); these percentages were similar. Also, regarding the subsequent chemotherapy regimen, the results for taxane were 48% vs. 21%, the results for platinum-based chemotherapy were 10% vs. 11%, and the results for fluoropyrimidine-based chemotherapy were 11% vs. 19%.

Given these data, the frequency of all-grade TRAEs in the nivolumab group was lower than that in the taxane group, indicating that nivolumab monotherapy might be safer than taxanes and has a different safety profile from that of taxanes. In terms of QOL, health-related QOL in the nivolumab arm was significantly better than that in the cytotoxic chemotherapy arm. However, serious irAEs, such as grade 4 diabetic ketoacidosis, occurred with nivolumab; hence, not only physical examinations, but also laboratory examinations such as the blood concentrations of hormones related to thyroid, pituitary and adrenal function should be performed.

4.2. Pembrolizumab

KEYNOTE-180 trial was phase 2, open-label, interventional, single-arm study, enrolled 121 patients [34]. Patients had advanced, metastatic esophageal cancer that progressed after 2 or more lines of therapy. Objective response rate was 9.9% (95% CI, 5.2%-16.7%) among all patients. In the KEYNOTE-180 trial, which was conducted as a phase II trial examining the use of pembrolizumab for EC patients, the incidences of all-grade TRAEs and grade 3–5 TRAEs were 57.9% and 12.4%, respectively. The most frequent all-grade TRAEs were fatigue (10.7%), rash (7.4%), pruritus (6.6%), diarrhea (5.8%) and pneumonitis (7.4%). Frequent grade 3–4 irAEs were type 1 diabetes mellitus (3.3%) and pneumonitis (2.4%). One patient (0.8%) had a grade 5 TRAE (pneumonia), and 5 patients (4.1%) discontinued pembrolizumab because of TRAEs (Table 3).

Table 3. Main outcomes and adverse events in the KEYNOTE-180 trial

	KEYNOTE-180
	Pembrolizumab
Number of patients	121
	ORR (%)	Median PFS (months)	Median OS (months)
All patients	9.9	2.0	5.8
SCC	14.3	2.1	6.8
AC	5.2	1.9	3.9
CPS≥10	13.8	2.0	6.3
CPS<10	6.3	2.0	5.4
Treatment-related AEs	Any grade	Grade ≥3
Any	70 (57.9)	15 (12.4)
Fatigue	13 (10.7)	0 (0)
Rash	9 (7.4)	0 (0)
Pruritus	8 (6.6)	0 (0)
Hypothyroidism	7 (5.8)	0 (0)
Diarrhea	7 (5.8)	1 (0.8)
Pneumonitis	9 (7.4)	3 (2.4)
Type 1 diabetes mellitus	4 (3.3)	4 (3.3)
Led to discontinuation	5 (4.1)
Led to death	1 (0.8)

Abbreviations: ORR, objective response rate; PFS, progression-free survival; OS, overall survival; SCC, squamous cell carcinoma; AC, adenocarcinoma; CPS, combined-positive score; AE, adverse event

As for the KEYNOTE-181 trial, TRAEs occurred in 64.3% of the patients in the pembrolizumab group and 86.1% of the patients in the chemotherapy group. The incidences of grade 3–5 TRAEs in the two groups were 18.2% and 40.9%, respectively. Patients in the cytotoxic chemotherapy group had higher incidences of TRAEs such as fatigue (20.6% vs. 11.8%), diarrhea (20.3% vs. 5.4%), nausea (21.6% vs. 7.0%), decreased appetite, (15.5% vs. 8.6%), alopecia (29.1% vs. 0.6%), peripheral sensory neuropathy (16.9% vs. 0.3%) and hematologic toxicities (white blood cell count decreased: 16.6% vs. 0.3%; neutropenia: 11.5% vs. 0%) in all grades, compared with patients in the pembrolizumab group. On the other hand, patients in the pembrolizumab group had higher incidences of TRAEs including hypothyroidism (11.5% vs. 2.4%), pneumonitis (4.8% vs. 0.7%), hyperthyroidism (4.1% vs. 0.7%) and hepatitis (1.9% vs. 0%). TRAEs led to treatment discontinuation in 6.1% and 6.4% of the patients in the pembrolizumab and cytotoxic chemotherapy groups, respectively, and no difference in deaths for the five patients in each group was seen (Table 4).

Table 4. Main outcomes and adverse events for the KEYNOTE-181 trial

		KEYNOTE-181
		Pembrolizumab		Chemotherapy
Number of patients		314		314
ORR (%)	All patients	13.1		6.7
	CPS≥10	21.5		6.1
	SCC	16.7		7.4
Median PFS (months)	All patients	2.1		3.4
	CPS≥10	2.6		3.0
	SCC	2.2		3.1
Median OS (months)	All patients	7.1		7.1
	CPS≥10	9.3		6.7
	SCC	8.2		7.1
AEs		N = 314 (100)		N = 296 (100)
		All grade	Grade ≥3	All grade	Grade ≥3
Anemia		8 (2.5)	4 (1.3)	66 (22.3)	23 (7.8)
White blood cell count decreased		1 (0.3)	0 (0)	49 (16.6)	30 (10.1)
Neutropenia		0 (0)	0 (0)	34 (11.5)	21 (7.1)
Diarrhea		17 (5.4)	2 (0.6)	60 (20.3)	9 (3.0)
Nausea		22 (7.0)	0 (0)	64 (21.6)	7 (2.4)
Decreased appetite		27 (8.6)	2 (0.6)	46 (15.5)	3 (1.0)
Fatigue		37 (11.8)	2 (0.6)	61 (20.6)	1 (0.3)
Alopecia		2 (0.6)	0 (0)	86 (29.1)	1 (0.3)
Peripheral sensory neuropathy		1 (0.3)	0 (0)	50 (16.9)	1 (0.3)
All events		202 (64.3)	57 (18.2)	255 (86.1)	121 (40.9)
Led to discontinuation		19 (6.1)		19 (6.4)
Led to death		5 (1.6)		5 (1.7)
Immune-mediated AEs		Pembrolizumab		Chemotherapy
Any		73 (23.2)		22 (7.4)
Hypothyroidism		36 (11.5)		7 (2.4)
Pneumonitis		15 (4.8)		2 (0.7)
Hyperthyroidism		13 (4.1)		2 (0.7)
Hepatitis		6 (1.9)		0 (0)
Colitis		3 (0.9)		2 (0.7)
Infusion-related reactions		3 (0.9)		8 (2.7)

Abbreviations: ORR, objective response rate; PFS, progression-free survival; OS, overall survival; SCC, squamous cell carcinoma; AC, adenocarcinoma; CPS, combined-positive score; AE, adverse event

The incidence of all-grade TRAEs in the pembrolizumab group was lower than that of all-grade TRAEs in the cytotoxic chemotherapy group, and pembrolizumab monotherapy might be a safer treatment than cytotoxic chemotherapy. However, similar to nivolumab, irAEs such as hypothyroidism, hepatitis, and myositis can occur with pembrolizumab treatment, so physical and laboratory examinations should be performed. Comparing the two anti-PD-1 antibodies such as nivolumab and pembrolizumab in terms of irAE, there seemed to be slightly difference of frequency. However especially in the case of endocrine AE, some cases show potential adverse events only detected by peripheral blood samplings without symptoms. At present, we think that there is not difference between these two ICIs for advanced EC patients. (Table 5)

Table 5. Any Grade of irAEs in four trials of anti-PD-1 antibody

	Nivo (AT-1) n=65	Nivo (AT-3) n=210	Pembro (KN-180) n=121	Pembro (KN-181) n=314
Diarrhoea	9 (13.8)	22 (10.5)	7 (5.8)	17 (5.4)
Nausea	-	4 (1.9)	-	22 (7.0)
Decreased appetite	6 (9.2)	16 (7.6)	-	27 (8.6)
Fatigue	5(7.7)	15 (7.1)	13 (10.7)	37 (11.8)
Hypothyrodism	-	2 (0.9)	7 (5.8)	36 (11.5)
Intestinal lung disease	3 (4.6)	5 (2.4)	-	-
Pneumonitis	-	4 (1.9)	9 (7.4)	15 (4.8)
Rash	5 (7.7)	23 (11.0)	9 (7.4)	-
Serious events	-	33 (15.7)	-	41 (20)
Events leading to death	0 (0)	2 (0.9)	1 (0.8)	5 (1.6)
Events leading to discontinuation	7 (10.8)	18 (8.6)	5 (4.1)	19 (6.1)

Abbreviations: Nivo, Nivolumab; Pembro, Pembrolizumab

4.3. Other immune checkpoint inhibitors

In China, ICIs other than nivolumab and pembrolizumab have been developed. Camrelizumab is a high-affinity, fully humanized, selective IgG4-κ monoclonal anti-PD-1 antibody. A randomized, open-label, ESCORT phase III trial was conducted at 43 centers in China[35]. In this trial, 457 ESCC patients were randomly assigned (1:1) to either a camrelizumab group (229 patients) or a chemotherapy group (docetaxel or irinotecan; 228 patients, respectively). The median OS was 8.3 months (95% CI: 6.8–9.7 months) in the camrelizumab group vs. 6.2 months (95% CI: 5.7–6.9 months) in the chemotherapy group (HR [95% CI]: 0.71 [0.57–0.87]; two-sided p = 0.0010), demonstrating that camrelizumab monotherapy showed a significant improvement compared with chemotherapy when used as a second-line treatment.

In the safety profiles, the incidences of TRAEs were similar in both the camrelizumab and chemotherapy groups (215 [94%] versus 198 [90%]). TRAEs of grade 3 or worse occurred in 44 (19%) of the 228 patients in the camrelizumab group and 87 (40%) of the 220 patients in the chemotherapy group. In the camrelizumab group, the most common grade 3 or higher adverse events were anemia (3%) and liver dysfunction (2%). Treatment-related deaths were reported for seven patients (three unknown causes, one enterocolitis, one hepatic dysfunction, one pneumonitis and one myocarditis) in the camrelizumab group and three patients (two unknown causes and one gastrointestinal hemorrhage) in the chemotherapy group. The most common irAEs were reactive capillary endothelial proliferation (182 [80%]), hypothyroidism (44 [19%]), skin reaction (20 [9%]), hepatitis (19 [8%]), pneumonitis (17 [7%]), and hyperthyroidism (14 [6%]). Reactive cutaneous capillary endothelial proliferation has been reported in other clinical trials of camrelizumab[36]. Reactive cutaneous capillary endothelial proliferation is characterized by a trilateral growth cycle of proliferation, plateau, and involution. It occurred as a grade 1–2 event (183 [80%]) or a grade 3 event (1 [<1%]), and most patients recovered after discontinuation.

In addition, a safety evaluation using patient-reported outcome and the EORTC QLQ-C30 and EORTC QLQ-OES18 was conducted, and valid responses were obtained from about 80% of all eligible patients. When health-related QOL in the camrelizumab arm was compared with QOL in the cytotoxic chemotherapy arm using the changes in EORTC QLQ-C30 and EORTC QLQ-OES18 scores between baseline and 8 weeks after treatment, the QOL evaluations (global health status, emotional functioning, social functioning, fatigue, pain, nausea and vomiting, appetite loss and diarrhea) performed using the EORTC QLQ-C30 showed significantly better results in the camrelizumab arm than in the cytotoxic chemotherapy arm, and QOL evaluations (reflux, taste problems and coughing problems) performed using the EORTC QLQ-OES18 also showed significantly better results in the camrelizumab arm than in the cytotoxic chemotherapy arm. Overall, the QOL in the camrelizumab arm tended to be maintained at a more general condition than the QOL in the cytotoxic chemotherapy arm for several items. The rates of subsequent therapy in the camrelizumab and cytotoxic chemotherapy arms in the ESCORT trial were 53.1% and 58.2%, respectively (radiotherapy: 14.0% vs. 19.5%, surgery: 0.9% vs. 0.5%, chemotherapy: 36.0% vs. 30.9%); these percentages were similar to the data obtained in the ATTRACTION-3 trial.

The ESCORT trial only included Chinese patients with ESCC patients, and Western and other Asian patients with ESCC were not examined; therefore, camrelizumab monotherapy is the first choice for second-line chemotherapy for metastatic or recurrent ESCC only in China. Additionally, the frequency and profile of grade ≥3 TRAEs of camrelizumab were quite similar to those of other immune checkpoint inhibitors such as nivolumab and pembrolizumab, which were lower than the frequencies of all-grade TRAEs in the cytotoxic chemotherapy group, and the rate of subsequent therapy in the camrelizumab group was similar to that in the cytotoxic chemotherapy group because of the maintenance of QOL.

5. Conclusion

ICIs such as nivolumab for advanced ESCC patients and pembrolizumab (PD-L1 CPS ≥10 and ESCC) for advanced EC patients showed survival benefits compared with cytotoxic agents; therefore, these ICIs have been established as standard second-line treatment options. Additionally, the frequencies of TRAEs in the ICI groups were less than those in the cytotoxic agent groups in the two phase III trials: ATTRACTION-3 trial and KEYNOTE-181 trial, suggesting that the safety of ICIs was manageable. Additionally, conventional cytotoxic drugs such as taxane or irinotecan might induce neutropenia, and infection in cases with aspiration pneumoniae as a result of stenosis can occasionally be fatal. With this background in mind, treatment with ICIs, which cause cytopenia less frequently, might be favorable especially for patients with metastatic or recurrent ESCC. However, although infrequent, ICIs can cause fatal irAEs, and physical examinations and metabolic endocrine-related blood tests should be performed before the administration of ICIs in clinical practice. Conventional cytotoxic agents such as taxane are key drugs for the treatment of patients with metastatic or recurrent EC, and particular attention to the presence of cytopenia or sensory neuropathy is needed in clinical practice. Also, since about 70% of metastatic or recurrent ESCC patients receive palliative radiation before salvage-line chemotherapy, the presence of ICI- or taxane-related pneumonia should be examined[5].

6. Expert opinions

Based on the ATTRACTION-3 trial and KEYNOTE-181 trial, ICIs have been established as standard treatment options, especially for patients with advanced ESCC, after platinum and fluoropyrimidine combination therapy [5,10]. However, not only ICI monotherapy as a second-line therapy, but also ICI-containing regimens have been developed for first-line or perioperative treatments in EC patients; thus, whether ICI-containing regimens, such as ICI plus platinum and fluoropyrimidine, ICI doublet, or preoperative ICI plus platinum and fluoropyrimidine, are tolerable for EC patients is an important question.

First, several phase III trials evaluating ICI-containing regimens as a first-line chemotherapy for metastatic or recurrent EC patients were conducted, and the KEYNOTE-590 phase III trial showed the superiority of pembrolizumab plus cisplatin and fluorouracil, compared with conventional cisplatin and fluorouracil, in terms of OS as a first-line chemotherapy for metastatic and recurrent EC patients in the ESMO 2020[37]. Although the safety data was limited, the frequencies of all-grade and grade ≥3 TRAEs in pembrolizumab plus cytotoxic chemotherapy group were reported to be 98.7% and 71.9%, respectively. By comparing these results with the safety data for the cytotoxic doublet chemotherapy group (all grades: 97.3%, grade ≥3: 67.6%), we expected a few significant differences in the frequency of TRAEs to exist as a result of the additional use of immune checkpoint inhibitor monotherapy. Regarding irAEs, the frequencies of all-grade and grade ≥3 TRAEs in pembrolizumab plus cytotoxic chemotherapy group were reported to be 25.7% and 7.0%, respectively. In the future, these safety profiles from the KEYNOTE-590 trial might provide basic data for the evaluation of other ICIs in development and cytotoxic agent combination therapy. Therefore, further detailed safety data are needed to evaluate this regimen in the KEYNOTE-590 trial.

There were few pivotal data regarding ICI-containing regimens as first-line treatment for metastatic or recurrent EAC patients. The CheckMate-649 phase III trial showed the superiority of nivolumab plus oxaliplatin and fluorouracil, compared with oxaliplatin and fluorouracil, in terms of OS as first-line treatment for metastatic and recurrent gastric cancer (GC) and gastroesophageal junction cancer (GEJC) and EAC patients in the ESMO 2020, although GC was 70% of patients in this study [38]. The frequencies of all-grade and grade ≥3 TRAEs in nivolumab plus cytotoxic chemotherapy group were reported to be 94% and 59%, respectively. The frequencies of all-grade and grade ≥3 TRAEs in conventional cytotoxic chemotherapy group were reported to be 89% and 44%, respectively. The total number of all-grade and grade ≥3 TRAEs regarding irAE were not presented, however the safety data of any adverse events was similar to these of the KEYNOTE-590 trial.

Secondly, in other cancers, such as advanced melanoma or lung cancer, ICI combination therapy such as nivolumab and ipilimumab, which is a human monoclonal antibody that targets CTLA-4, showed efficacy as a first-line chemotherapy [39,40]. However, although ICI combination therapy was reportedly effective, the frequency of irAEs was high. For metastatic esophageal and oesophagogastric and gastric adenocarcinoma patients, the CheckMate-032 phase I/II trial evaluating the safety and efficacy of doublet therapy consisting of nivolumab and ipilimumab showed that the objective response rates were 24% for the nivolumab (1 mg/kg) plus ipilimumab (3 mg/kg) group and 8% for the nivolumab (3 mg/kg) plus ipilimumab (1 mg/kg) group[41]. In the safety profiles, serious TRAEs occurred in 43% of the patients in the nivolumab (1 mg/kg) plus ipilimumab (3 mg/kg) group and 25% in the nivolumab (3 mg/kg) plus ipilimumab (1 mg/kg) group. In particular, diarrhea (8%) and adrenal insufficiency (8%), fatigue (6%), increased ALT (6%), increased AST (6%), and colitis (6%) were frequent serious TRAEs in the nivolumab (1 mg/kg) plus ipilimumab (3 mg/kg) group, and pneumonitis (8%) was a frequent serious TRAE in the nivolumab (3 mg/kg) plus ipilimumab (1 mg/kg) group.

Given these data, nivolumab plus ipilimumab could be a promising therapy for metastatic or recurrent EC in terms of efficacy; in a cross-trial comparison, the frequency of serious irAEs in ICI combination therapy seemed to be higher than that in ICI monotherapy, such as nivolumab or pembrolizumab. A CheckMate-648 phase III trial comparing nivolumab plus cisplatin and fluorouracil or nivolumab plus ipilimumab with cisplatin and fluorouracil as first-line chemotherapy for untreated patients with metastatic or recurrent ESCC is released in the ASCO 2021[42]. The trial showed the superiority of nivolumab plus cisplatin and fluorouracil group and nivolumab plus ipilimumab group compared to conventional cisplatin and fluorouracil group, regarding OS in PD-L1 (TPS) ≥ 1% and all populations.

Following the result of the KEYNOTE-590 trial and the CheckMate-648 trial, ICI plus platinum and fluoropyrimidine regimen and ICI doublet regimen might be standard treatments as first-line treatment for metastatic or recurrent EC patients in the near future. Whether there are no data on efficacy comparison of these two regimens in the CheckMate-648 trial, safety profiles of two regimens are important to select in clinical practice. Although the safety data was limited, the frequencies of all grade and grade ≥3 TRAEs in nivolumab plus cytotoxic chemotherapy group were reported to be 96% and 47%, respectively. On the other hand, the frequencies of all grade and grade ≥3 TRAEs in nivolumab plus ipilimumab group were reported to be 80% and 32%, respectively. The frequencies of treatment-related death were 2% each group. In terms of irAEs, the frequencies of all grade and grade ≥3 TRAEs in nivolumab plus cytotoxic chemotherapy group were reported to be 6–24% and 1–2% for each organ function, respectively. On the other hand, the frequencies of all grade and grade ≥3 TRAEs in nivolumab plus ipilimumab group were reported to be 2–35% and 1–6% for each organ function, respectively. Considering these safety data, frequencies of some irAEs such as endocrine, hepatic, skin in nivolumab plus ipilimumab group seem to be slightly higher than in nivolumab plus chemotherapy group. But no statistical difference was shown between two groups in the report, it might be difficult to select these regimens only based on safety data. Therefore we need to take not only safety data but also patients’organ functions such as kidney (intolerant to cisplatin) into account for selecting nivolumab plus ipilimumab or nivolumab plus doublet chemotherapy for advanced ESCC patients.

Thirdly, ICI-containing preoperative chemotherapy is being developed for various cancer types, such as lung and breast, and some studies have shown promising efficacy and careful safety data. For early-stage breast cancer patients, the I-SPY2 phase II trial, which evaluated the efficacy and safety of preoperative treatment consisting of pembrolizumab plus cytotoxic agents, showed promising results for a pathological complete response. On the other hand, adrenal insufficiency was reported in 6 out of 69 patients (8.7%), and 5 of these patients continued to exhibit adrenal insufficiency after surgery[43]. In the ATTRACTION-3 and KEYNOTE-181 trials, only one patient and no patients showed adrenal insufficiency, respectively. Similarly, for resectable locally advanced ESCC patients, the JCOG1804E (FRONTiER) phase I trial, which evaluated the safety of preoperative treatments consisting of nivolumab plus cisplatin and fluorouracil, showed that one patient in the cohort (all 7 patients) had grade 2 adrenal insufficiency after surgery[4445]. In postoperative case, a CheckMate-577 trial showed the efficacy of adjuvant nivolumab in resected EC and GEJC following neoadjuvant chemoradiotherapy and the frequencies of all-grade and grade ≥3 TRAEs in terms of endocrinal adverse event were 17% and <1%, respectively in this trial [45]. Based on these safety profiles for preoperative ICI-containing therapy, we are concerned that surgical invasion might affect the onset of adrenal insufficiency. However, few reports describing the precise safety profiles of ICI-containing preoperative treatments are presently available, and further investigations are needed.

Article highlights

• The prognosis of advanced EC is poor, and few treatment options exist for metastatic or recurrent EC.

• Immune checkpoint inhibitors such as nivolumab and pembrolizumab, as well as camrelizumab, have shown clinical benefits, compared with cytotoxic agents, as second-line treatments for oesophageal squamous cell carcinoma.

• When selecting conventional cytotoxic agents such as paclitaxel, docetaxel and irinotecan, care is needed with regard to the different treatment schedules and the different safety profiles, such as neutropenia, febrile neutropenia, sensory neuropathy, anorexia and diarrhoea. Therefore, when adverse events occur, the dose reduction of cytotoxic agents, treatment discontinuation, or additional supportive care might be required.

• Immune checkpoint inhibitors are associated with lower frequencies of treatment-related adverse events than conventional cytotoxic agents. But in some cases, serious immune-related adverse events such as pneumonia, fulminant type 1 diabetes and myocarditis have been reported. When immune-related adverse events occur, the administration of immune checkpoint inhibitors should be discontinued; the addition of appropriate hormone replacement therapy or the use of corticosteroid might also be necessary.

Declaration of interest

K Kato has declared research funds from ONO, MSD, Merck Bio, BMS, Taiho, Chugai, Oncolys Biopharma and Shionogi. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.