Nivolumab for esophageal squamous cell carcinoma and the predictive role of PD-L1 or CD8 expression in its therapeutic effect

Abstract

Introduction

Nivolumab, a programmed death 1 (PD-1) inhibitor, has recently demonstrated efficacy as second-line therapy for esophageal squamous cell carcinoma (ESCC) patients in a phase III trial. We report real-world clinical outcomes of nivolumab therapy for ESCC patients.

Methods

ESCC patients refractory/intolerant to at least one line of chemotherapy and who received nivolumab as a subsequent line of therapy were included. The efficacy and safety of nivolumab and the predictive role of PD-L1 and CD8 expression were analyzed.

Results

Fifty-eight patients were analyzed for safety and survival outcomes, while 57 were analyzed for objective response rates (ORR) excluding one with no measurable lesions. Eleven patients achieved a partial response, leading to an ORR of 19.3%. The median response duration was 6.5 months (range 4.1–22.4). The median progression-free survival (PFS) and overall survival were 2.1 (95% confidence interval [CI] 1.8–2.3) and 7.4 (95% CI 4.8–10.0) months, respectively. Among patients with adequate samples, 56.9% (29/51), 27.5% (14/51), and 17.6% (9/51) expressed a combined positive score (CPS) ≥ 1, ≥ 10, and ≥ 20, respectively, while 24.4% (11/45) and 57.5% (23/40) were positive for intratumoral and peritumoral CD8 + T cell infiltration, respectively. A significantly longer PFS was observed in patients with a CPS ≥ 20 (7.5 [95% CI 1.8–13.1] vs. 1.9 [1.4–2.3] months, P = 0.05), and a trend towards better survival was seen in those with CPS ≥ 10 or intratumoral CD8 + T cell infiltration.

Conclusions

Nivolumab is a valuable option at subsequent treatment lines for patients with advanced ESCC.

Electronic supplementary material

The online version of this article (10.1007/s00262-020-02766-7) contains supplementary material, which is available to authorized users.

Introduction

Esophageal cancer ranked seventh in incidence (572,000 new cases) and sixth in mortality (509,000 deaths) worldwide in both sexes in 2018 [1]. Esophageal cancer consists mostly of two main histologic types: esophageal adenocarcinomas (ECAs) and squamous cell carcinomas (ESCCs).

While ESCC has been decreasing in incidence in North America and West Europe, it remains the highly prevalent histologic type of esophageal cancer in East Asia and Africa [2]. Among 2483 patients with newly diagnosed esophageal cancer in South Korea in 2017, 2255 (90.8%) cases were ESCC and 60 (2.4%) were ECAs [3]. This epidemiologic characteristic was similar to that from other East Asian countries where ESCC comprised over 90% of total cases [2, 4].

Cisplatin in combination with fluorouracil in first-line therapy, followed by taxane or irinotecan in second-line therapy, constitutes the standard treatment for advanced esophageal cancer. However, sequential chemotherapy results in disappointing outcomes with a median overall survival of less than 1 year [5].

Immune checkpoint inhibitors have revolutionized the treatment schemes across many cancer types since their first approval for melanoma in 2014 [6]. Anti-programmed cell death protein 1 (PD-1) antibodies inhibit the interaction between the PD-1 expressed on T cells and its ligands, programmed cell death ligand-1 (PD-L1) and programmed cell death ligand-2, expressed on antigen-presenting cells and tumor cells, thereby reverting the ability of T cells to recognize and kill tumor cells [7].

As expected, efforts to evaluate the efficacy of immunotherapy have expanded to include esophageal cancer. Pembrolizumab, a PD-1 inhibitor, proved its efficacy as second-line therapy for esophageal cancer in KEYNOTE-181 [8], and has been approved in the USA since August 2019 for patients with high PD-L1 expressing esophageal cancer.

Nivolumab, another well-known PD-1 inhibitor, was studied in a single-arm phase II trial (ATTRACTION-1) of 65 Japanese patients with ESCC [9]. The drug showed a promising efficacy with acceptable toxicity profiles in ESCC patients who had progressed after chemotherapy. Another landmark study (ATTRACTION-3) compared the efficacy of second-line nivolumab to that of conventional chemotherapy for ESCC patients [10], and successfully demonstrated an overall survival benefit in those treated with nivolumab. Based on these results, since February 2020, nivolumab has been approved in Japan for patients with ESCC that had progressed following chemotherapy.

In Korea, nivolumab was not endorsed for the treatment of esophageal cancer until official approval was granted in April 2020. However, based on the outcome of the ATTRACTION-1 trial, our institute received preliminary approval in November 2017 to use nivolumab in the treatment of patients whose cancer had progressed after fluoropyrimidine-, platinum-, or taxane-based chemotherapy. Here, we present the outcomes of nivolumab, with comprehensive analysis on the predictive role of PD-L1 and CD8 expressions, as second or further-line therapy for ESCC patients.

Materials and methods

Patients and data collection

We conducted a retrospective analysis on patients diagnosed with ESCC who were treated with nivolumab in second- or further-line therapy from November 2017 to November 2019. All patients were pathologically confirmed to have ESCC and refractory or intolerant to at least one line of fluoropyrimidine-, platinum- or taxane-based chemotherapy.

Nivolumab therapy

Patients received 3 mg/kg nivolumab intravenously every 2 weeks until disease progression, unacceptable toxic effects, or death. The Response Evaluation Criteria in Solid Tumors version 1.1 were used to assess the treatment response [11]. Adverse events (AEs) were assessed according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), version 4.0.

Immunohistochemistry (IHC)

PD-L1 expression was assessed in pre-treatment samples obtained from endoscopic biopsies or surgical resection. Assessment of PD-L1 expression was done by experienced pathologists at our institution. All samples except one, which was tested with PD-L1 22C3 pharmDx, were tested using a PD-L1 IHC SP263 kit on the DAKO ASL48 platform (Tucson, AZ) according to the manufacturer’s recommendations. The combined positive score (CPS) was calculated by summing the number of PD-L1-stained cells (tumor cells, lymphocytes, and macrophages) and dividing the result by the total number of viable tumor cells, multiplied by 100. The tumor proportion score (TPS) was calculated by summing the number of PD-L1-stained tumor cells and dividing the result by the total number of viable tumor cells, multiplied by 100 (supplementary Fig. 1) [12].

For the CD8 IHC test, paraffin-embedded tumor sections were dewaxed in xylene and ethanol, and autoclaved for 60 min in an antigen retrieval solution (CC1), and were incubated for 24 min in Ventana BenchMark XT with anti-CD8 (RTU; clone SP57; Ventana Medical Systems, Roche, Tucson, Arizona, USA). The secondary antibody was incubated with an ultraview universal DAB detection kit (Ventana Medical Systems, Roche, Tucson, Arizona, USA). A PD-L1 IHC test using a PD-L1 IHC 22C3 pharmDx kit on the DAKO ASL48 platform was performed according to the manufacturer’s recommendations.

CD8 + T cells were quantified by screening the representative cancer area using a modification of the previous methods as follows [13]. For the quantitative evaluation of intratumoral and peritumoral CD8 + T cells in representative sections, we chose three microscopic fields, counted intratumoral/peritumoral CD8 + T cells at × 200 magnification, and averaged them respectively (supplementary Fig. 2). CD8 + T cell infiltration was considered positive when the average number of CD8 + T cells was > 50 [13].

Statistical analysis

All patients were included for survival analysis, and only those with measurable lesions were included for response analysis. The comparison of objective response rates (ORR) according to specific PD-L1 expression levels was performed using the Fisher’s exact test or χ² test. Progression-free survival (PFS) was defined as the time from initiation of nivolumab to the documentation of either disease progression or death. Overall survival was defined as the time from the initiation of nivolumab to death from any cause. Survival curves were calculated using the Kaplan–Meier method and compared using the log-rank test. Patients were censored at the time of analysis if they were progression-free and/or alive. All analyses were performed using IBM SPSS Statistics 25 (Armonk, NY), and a P value of less than 0.05 was considered statistically significant. The data cutoff date was January 31, 2020.

Results

Patient characteristics

A total of 58 patients diagnosed with ESCC who had received at least one cycle of nivolumab as second- or further-line therapy were included in this study. Table 1 shows patient characteristics. The median age was 65 (range 39–85) and most patients were male and had good Eastern Cooperative Oncology Group (ECOG) performance status (PS), while five patients had ECOG PS 2 or 3. Fifty percent of patients received one prior chemotherapy, whereas 37.9% received two, and 12.1% received three or more. Additionally, 44.8% and 31.0% underwent surgery and concurrent chemoradiation, respectively, with curative intent.

Table 1.

Baseline characteristics

Characteristics	N = 58 (%)
Age (range), years	65 (39–85)
Sex
Male	54 (93.1)
Female	4 (6.9)
ECOG performance status
1	53 (91.4)
≥ 2	5 (8.6)
Histologic type
Squamous cell carcinoma	58 (100.0)
Alcohol consumption
Ever	48 (82.8)
Never	10 (17.2)
Smoking
Ever	52 (89.7)
Never	6 (10.3)
Previous treatment
Curative surgery	26 (44.8)
Concurrent chemoradiation	18 (31.0)
Number of prior chemotherapies
1	29 (50.0)
2	22 (37.9)
≥ 3	7 (12.1)
PD-L1 expression (n = 51)
TPS
≥ 1%	16 (31.4)
≥ 10%	9 (17.6)
≥ 20%	7 (13.7)
CPS
≥ 1	29 (56.9)
≥ 10	14 (27.5)
≥ 20	9 (17.6)

ECOG Eastern Cooperative Oncology Group, PD-L1 programmed cell death ligand-1, TPS tumor proportion score, CPS combined positive score

PD-L1 and CD8 expression

Of the 58 patients included in this study, 51 had samples suitable for the assessment of PD-L1 expression. Based on TPS, 16 (31.4%) patients were determined to be PD-L1-positive (TPS ≥ 1%), including 9 (17.6%) and 7 (13.7%) patients with a TPS ≥ 10% and ≥ 20%, respectively. Based on CPS, 29 (56.9%) patients were determined to be PD-L1-positive (CPS ≥ 1), including 14 (27.5%) and 9 (17.6%) patients with a CPS ≥ 10 and ≥ 20, respectively.

In the CD8 IHC test, 45 and 40 patients had adequate samples for evaluation of IT and PT microscopic fields, respectively: 5 samples were suitable for only IT evaluation, because the representative microscopic fields from tumor samples obtained by small biopsies did not contain an adequate number of PT fields. IT CD8 + T cell infiltration was positive in 11 of 45 (24.4%) samples, and PT CD8 + T cell infiltration was positive in 23 of 40 (57.5%) samples.

Efficacy of nivolumab and predictive role of PD-L1 or CD8 expression

A total of 57 patients, excluding one with no measurable lesion at the baseline image, were included in the response analysis. Among them, six patients were lost to a follow-up imaging evaluation due to the aggravation of cancer-related symptoms. Although their diseases were most probably suspected to be non-responsive to nivolumab, they were included in the analysis of ORR as being an unevaluable objective response. Finally, 11 of 57 patients achieved a partial response with an ORR of 19.3% (11/57) (Table 2). Eleven (19.3%) additional patients achieved stable disease.

Table 2.

Treatment efficacy

	N = 58 (%)
Best response (n = 57)a
Partial response	11 (19.3)
Stable disease	11 (19.3)
Progressive disease	29 (50.9)
Not evaluable	6 (10.5)
Objective response rate	11 (19.3)
Median duration of treatment (range), months	1.9 (0.5–25.4)
Median time to response (range), months	1.6 (1.0–3.2)
Median duration of response (range), months	6.5 (4.1–22.4)
Discontinued treatment	47 (81.0)
Reason for treatment discontinuation (n = 47)b
Disease progression	39 (83.0)
Symptomatic deterioration	4 (8.5)
Transfer to another hospital	1 (2.1)
Follow-up loss	3 (6.4)

^aOne patient with only non-measurable lesions, who showed non-CR/non-PD, was excluded from the response analysis

^bOf 48 patients who showed a disease progression according to RECIST v1.1, one patient maintained clinical benefit and continued with nivolumab treatment

Although not statistically significant due to the small number of patients analyzed, patients with higher PD-L1 expression levels (1%, 10%, and 20% on TPS and 1, 10, and 20 on CPS) tended to show a higher response rate than those with lower PD-L1 expression (supplementary Table 1; N = 51 for patients with PD-L1 tested). However, even patients with PD-L1-negative tumors according to CPS, demonstrated an ORR of 18.2% (4 of 22). The waterfall plot (Fig. 1; N = 49 patients with baseline measurable lesions and at least one imaging evaluation after treatment initiation) additionally showed a consistent response depth regardless of PD-L1 positivity analyzed according to CPS.

Fig. 1.

Waterfall plot of best percentage change for target lesions by RECIST v1.1. PD-L1 status determined by CPS are indicated (purple/plain: positive, green/horizontal lines: negative, black/X: not evaluated). CPS combined positive score

The median duration of therapy for the total study population (N = 58) was 1.9 months (range 0.5–25.4). For 11 patients with a partial response, the median time to initial response was 1.6 months (range 1.0–3.2) and the median duration of the response was 6.5 months (range 4.1–22.4). The swimmer plot (N = 58) in Fig. 2 shows that patients demonstrated a durable response irrespective of PD-L1 CPS positivity. One patient, whose tumor expressed PD-L1 CPS 5 and TPS 0%, showed a dramatic tumor shrinkage maintained over two years (Fig. 3 the patient represented by the bar ranked highest in Fig. 2).

Fig. 2.

Swimmer plot of treatment duration. PD-L1 status determined by CPS are indicated (purple/plain: positive, green/horizontal lines: negative, black/X: not evaluated). CPS combined positive score

Fig. 3.

Contrast-enhanced chest CT scan of the patient with the longest duration of response at baseline, 3 months, 13 months, and 24 months after treatment initiation. Arrowhead, left lower paratracheal lymph node. Arrow, paraesophageal lymph node

The median follow-up duration for overall survival was 13.8 months. For the entire study population (N = 58), the median PFS was 2.1 months (95% CI 1.8–2.3) and the median overall survival was 7.4 months (95% CI 4.8–10.0) (supplementary Fig. 3). The comparison of survival outcomes according to PD-L1 expression status indicated a trend, albeit not significant, toward better PFS in patients with a PD-L1 CPS ≥ 1 (2.4 [95% CI 1.4–3.4] vs. 1.6 [0.8–2.4], P = 0.20) (Fig. 4a), a CPS ≥ 10 (2.6 [1.1–4.2] vs. 2.0 [1.5–2.6] months, P = 0.24) (Fig. 4b), a TPS ≥ 10% (3.3 [1.3–5.4] vs. 2.0 [1.5–2.6] months, P = 0.15) (Fig. 4d), and a TPS ≥ 20% (7.5 [0.0–15.9] vs. 2.0 [1.6–2.5] months, P = 0.069) (Fig. 4f). A significant difference in PFS was seen among patients with a CPS ≥ 20 (7.5 [1.8–13.1] vs. 1.9 [1.4–2.3] months, P = 0.05) (Fig. 4c). The 12-month PFS rate was 38% for those with a CPS ≥ 20 and 11% for those with a CPS < 20 (Fig. 4c); and 38% for those with a TPS ≥ 20% and 10% for those with a TPS < 10% (Fig. 4f).

Fig. 4.

Kaplan–Meier plots of progression-free survival according to PD-L1 expression levels at a a CPS of 1, b a CPS of 10, c a CPS of 20, d a TPS of 1%, e a TPS of 10%, and f a TPS of 20%. CI confidence interval, CPS combined positive score, HR hazard ratio, TPS tumor proportion score

Response rates and survival outcomes did not differ significantly according to CD8 + T cell infiltration (supplementary Table 2, supplementary Fig. 5). However, patients with a PD-L1 CPS ≥ 10 or positive for IT CD8 + T cell infiltration showed a numerically longer survival duration (PFS: 2.4 vs. 1.7 months, P = 0.34; OS: 7.5 vs. 5.5, P = 0.41) and nearly two-fold ORR (24.0 vs. 10.5%) and disease control rate (44.0 vs. 21.1%) compared to those whose tumor was negative for both markers (Fig. 5).

Fig. 5.

Tumors with a PD-L1 CPS ≥ 10 or that were positive for IT CD8 + T cell infiltration showed a a higher overall response rate (in green) and disease control rate (in blue), b longer progression-free survival and c overall survival. CPS combined positive score, DCR disease control rate, IT intratumoral, PD-L1 programmed death ligand 1, PFS progression-free survival, ORR overall response rate, OS overall survival

Safety of nivolumab

There were 11 treatment-related AEs in 10 patients: 8 cases of hypothyroidism (grade 2), 2 cases of pneumonitis (two grade 1), and 1 colitis (grade 3). Treatment discontinuation was required for only one patient who developed grade 3 colitis. Two patients with grade 1 pneumonitis presented no symptoms but were diagnosed by ground-glass infiltrates on scheduled imaging evaluation, while maintaining a partial remission status. Both patients continued with nivolumab treatment and were closely followed-up for any symptom or imaging aggravation.

Discussion

In this study, we confirmed the clinical efficacy of nivolumab as second- or further-line therapy for ESCC from real-world data. The efficacy values from our study are compatible with those from previous prospective studies [8–10]. The ORR of 19.3% that we obtained is very similar to that of 17.2% or 21.9% (centrally or investigator-assessed, respectively) from ATTRACTION-1, 16.7% from the ESCC subgroup of KEYNOTE-181, and 19.3% from ATTRACTION-3. However, the median overall survival of 7.4 months was slightly shorter than those from prior prospective trials ranging from 8.2 to 10.9 months.

Many factors contribute to the relatively shorter duration of overall survival in our study. First, five patients (9%, 5 of 58) who had ECOG PS 2 or 3 when they first received nivolumab were included in our study, reflecting the patient subgroup encountered in the real-world clinic. These five patients showed a very short overall survival duration of 3.2 months. Poor PS has been previously reported as a negative predictive marker for survival outcome of immunotherapy [14]. In comparison to our study, the prior three prospective studies included only patients with a good PS of 0 or 1 [8–10].

Second, only half of our study population received nivolumab as second-line therapy, while the other half received nivolumab as third-, fourth- or fifth-line therapy. In comparison, all patients enrolled in ATTRACTION-3 and KEYNOTE-181 studies were administered nivolumab and pembrolizumab, respectively, as second-line therapy only [8, 10]. In the subgroup analysis not presented in our Results section, the median overall survival differed according to line of therapy: 7.8 months for patients who received nivolumab as second-line therapy, 6.0 months for those receiving it as third-line therapy, and 3.2 months for those receiving it as fourth- or fifth-line therapy (P = 0.011).

Two phase III trials showed inconclusive results regarding the predictive role of PD-L1 expression in the efficacy of PD-1 inhibitors [8, 10]. Although the whole study population in KEYNOTE-181 with PD-L1 CPS ≥ 10, which consisted of ESCC and EAC, showed a superior overall survival when treated with pembrolizumab compared to chemotherapy, this survival benefit was not statistically significant for the SQCC subgroup [8]. On the other hand, while the ATTRACTION-3 study demonstrated nivolumab treatment reduced the risk of death in patients with PD-L1 TPS ≥ 1%, the actual overall survival duration did not differ between the two groups. Furthermore, there was no significant interaction between nivolumab treatment efficacy and the status of PD-L1 expression [10]. However, the ATTRACTION-3 did not report clinical outcomes according to CPS or other cutoff values by TPS.

To determine whether the efficacy of nivolumab is more predictable by CPS or other cutoff values by TPS, we measured the expression of PD-L1 using both TPS and CPS with various cut-off levels. Our data suggest that a CPS cutoff value of 20 predicts a significant PFS benefit of nivolumab treatment, while a CPS/TPS cutoff value of 10% was more predictable for treatment benefit compared to the cutoff value of 1%. However, the differentiating power of PD-L1 expression was minimal for overall survival.

Previous PD-1/PD-L1 trials across many tumor types demonstrated that the predictive role of PD-L1 expression, if any, is more prominent for overall survival than for PFS [15–17]. In our study, however, PD-L1 expression failed to show any clinically meaningful difference in overall survival. This lack of prediction may be attributable to the large proportion of PD-L1-negative patients in our study population who received post-nivolumab chemotherapy. Fifty-five percent of patients (26 of 47 patients who stopped nivolumab) in our study received post-nivolumab chemotherapy, which was a similar proportion to that reported in the ATTRACTION-3 study [10]. However, when we analyzed these patients according to PD-L1 status, 45% (10/22) belonged to PD-L1-positive, 68% (13/19) to PD-L1-negative, and 50% (3/6) to the PD-L1-unknown group by CPS. Prevalent use of post-nivolumab chemotherapy in the PD-L1-negative group may have improved post-progression survival, leading to attenuation of the predictive effect of PD-L1 expression.

Additionally, we evaluated the role of CD8 + T cell infiltration as a predictive or prognostic marker for nivolumab treatment in ESCC. CD8 + T cells are an important component of tumor-infiltrating lymphocytes, which manifest as a host immune response against cancer cells. However, there are several contrasting reports on the prognostic or predictive roles of CD8 + T cells [18, 19]. Our study showed that infiltration of CD8 + T cells can be used adjunctively with PD-L1 expression as a predictive marker. Though this issue should be further investigated, the lack of predictive power of CD8 + T cell infiltration alone is supported by a previous study, which reported that the expression of human leukocyte antigen (HLA) class I, a prerequisite molecule for activating CD8 + T cells, is deficient in ESCC, suggesting that some of these CD8 + T cells are inactive in the anti-tumor immune response [20].

Biomarker analysis using PD-L1 and CD8 + T cells demonstrated that PD-L1 expression with or without supplementary CD8 + T cell infiltration has a predictive role for nivolumab therapy. However, the predictive power was not sufficient to exclude nivolumab as a treatment option for patients with weak or negative PD-L1 expression or CD8 + T cell infiltration. Further search is needed to identify novel and reliable biomarkers.

There are several limitations to this study. Due to its retrospective nature, our study included a heterogeneous population with regard to PS and treatment line. Therefore, it was difficult to evaluate the long-term effect of nivolumab on overall survival, although efficacy regarding response rate and PFS, which require a relatively shorter duration of follow-up, was sufficiently analyzed. On the other hand, our use of real-world data represents the patient population encountered in the clinic and provides evidence for the role of nivolumab as second- or further-line therapy for patients with advanced ESCC.

In conclusion, our study suggests that nivolumab is a promising option in subsequent treatment lines, with an acceptable toxicity profile for patients with advanced ESCC. This drug should be considered for ESCC patients who experience progression after first- or second-line chemotherapy.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contribution

JL, YLC, and JMS contributed to the study conception and design. Material preparation, data collection and analysis were performed by JL, BK, HAJ, YLC, and JMS. The first draft of the manuscript was written by JL and JMS, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by the National R & D Program for Cancer Control, Ministry of Health & Welfare, Korea [1720180].

Availability of data and material

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Code availability

Not applicable.

Compliance with ethical standards

Conflict of interest

The authors have no conflict of interest to declare.

Ethics approval

This study was approved by the Institutional Review Board of Samsung Medical Center with a waiver for informed consent in view of the retrospective nature of the study and all the procedures being performed were part of the routine care (IRB No. 2020-03-087).

Consent to participate

Not applicable.

Consent to publish

Not applicable.