Adding anti-PD-1 antibody to definitive chemoradiotherapy in elderly patients with esophageal squamous cell carcinoma: higher intensity does not equate to better outcomes

Abstract

Background and purpose

The benefit of adding anti-PD-1 antibodies to definitive chemoradiotherapy (dCRT) in elderly patients with esophageal squamous cell carcinoma (ESCC) remains uncertain. This study evaluated its efficacy and safety versus dCRT alone.

Materials and methods

We retrospectively analyzed the patients aged ≥ 70 years with ESCC treated at three academic centers from 2009 to 2023. All patients received first-line dCRT and the study group additionally received anti-PD-1 antibodies (IO group). Propensity score matching (PSM) was applied to balance baseline factors.

Results

A total of 241 patients were enrolled, including 130 in the IO group and 111 in the dCRT group. After 1:1 PSM (110 patients per group), no significant differences in overall survival (OS) or progression-free survival (PFS) were observed. The median OS was 34.5 vs 33.7 months (HR = 0.86, 95%CI: 0.58–1.28, p = 0.467) and median PFS was 29.8 vs 17.8 months (HR = 0.79, 95%CI: 0.55–1.13, p = 0.194). Multivariate Cox analysis identified high nutritional risk as an independent predictor of worse OS (p = 0.014), while both advanced TNM stage (p = 0.030) and high nutritional risk (p = 0.016) were independently associated with shorter PFS. Subgroup analyses suggested that patients with good performance, better nutritional status or lower comorbidity burden may benefit from combination therapy. Grade 3–4 adverse events were comparable between two groups.

Conclusion

Adding anti-PD-1 antibodies to dCRT did not result in a significant improvement in OS or PFS in the ESCC patients aged ≥ 70 years; however exploratory findings indicate a potential PFS signal in selected patients with favorable baseline conditions, which requires confirmation in prospective studies.

KEY MESSAGES

1. This is the first multicenter study to evaluate the addition of immunotherapy to definitive chemoradiotherapy in elderly patients with esophageal squamous cell carcinoma.

2. The addition of immune checkpoint inhibitors did not improve survival outcomes in the overall older population.

3. Nutritional status and performance condition were strong prognostic factors; elderly patients with favorable profiles may derive greater benefit from intensified treatment.

Introduction

Esophageal cancer (EC) represents a significant global health burden, with China accounting for over 50% of cases worldwide. Approximately 85% of Chinese EC cases are squamous cell carcinoma (ESCC), and 68.9% of patients are diagnosed after age 60, including 40% aged ≥ 70 years [1–5]. Despite therapeutic advances, the 5-year survival rate for locally advanced disease remains < 25% [4], highlighting the urgent need for more effective treatments.

Elderly ESCC patients present unique management challenges. Although surgical resection remains the primary curative option [6], age-related physiological decline and frequent comorbidities often preclude surgery. Definitive chemoradiotherapy (dCRT) has become the standard non-surgical treatment since the RTOG85-01 trial demonstrated superior survival compared with radiotherapy alone (5-year survival: 26% vs 0%) [7]. However, toxicity remains a concern in older adults, with 60-70% experiencing grade ≥ 3 adverse events [7]. Modified regimens, such as oral S-1 chemotherapy with radiotherapy, have shown efficacy in elderly patients, yielding superior 2-year OS (53.2% vs 35.8%, HR = 0.63, p = 0.002) and complete response rates (41.6% vs 26.8%, p = 0.007) with acceptable toxicity [8], consistent with another trial reporting improved 3-year OS (46.2% vs 33.9%, p = 0.02) [9]. Nevertheless, these outcomes remain suboptimal, with no significant reduction in distant metastasis (3-year distant metastasis-free survival: 41.2% vs 32.3%, HR = 0.76, 95%CI: 0.58–1.00, p = 0.50) [9], underscoring the need for novel systemic approaches.

The treatment landscape for advanced ESCC has been transformed by several phase III clinical trials [10–14], consistently showing that the adding anti-PD-1 antibodies to chemotherapy significantly prolongs OS compared with chemotherapy alone. Median OS improvements typically range from 12 to 17 months versus 10 to 12 months, with hazard ratios generally between 0.58 and 0.74 (all p < 0.05). For locally advanced ESCC, the EC-CRT-001 trial [15] achieved a complete response rate of 62% with toripalimab plus dCRT. Importantly, these trials enrolled limited elderly cohorts: KEYNOTE-590 had a median age of 64 (range 28–94) [10], while JUPITER-06 and EC-CRT-001 reported median ages of 63 (range 20–75) and 56 (IQR 53–63) years, respectively [11,15]. Similarly, major trials such as CheckMate-648(12), ESCORT-1st [13], and ORIENT-15(14) included adults ≥ 18 years, but none provided dedicated analyses for geriatric populations, leaving gaps in understanding the efficacy and safety of immunotherapy in elderly patients.

Given immunotherapy’s proven efficacy in advanced ESCC and emerging potential in locally advanced disease, as well as its theoretically favorable safety profile in older adults, we investigated the efficacy and safety of anti-PD-1 antibodies combined with dCRT specifically in elderly (≥ 70 years) ESCC patients, aiming to optimize therapeutic strategies for this vulnerable population.

Methods and materials

Patient population

We retrospectively collected data from patients diagnosed with ESCC between March 2009 and December 2023 at three academic cancer centers (Sun Yat-sen University Cancer Center, Zhejiang Cancer Hospital and Hubei Cancer Hospital). The data were analyzed in January 2025. Eligible patients met the following criteria: (1) histologically confirmed ESCC; (2) age ≥ 70 years at diagnosis; (3) clinical stage TanyNanyM0 (AJCC 8th edition) or M1 limited to supraclavicular lymph nodes; (4) received definitive radiotherapy (≥50 Gy) and chemotherapy, with or without anti-PD-1 therapy; (5) patients in the IO group received at least two cycles of immunotherapy. Exclusion criteria included: (1) age < 70 years; (2) palliative surgery; (3) chemotherapy used solely as adjuvant treatment; (4) concurrent other malignancies; (5) incomplete medical records. Comorbidities prior to ESCC diagnosis were assessed using the Charlson Comorbidity Index (CCI) [16], and all patients underwent nutritional risk screening at admission using the Nutritional Risk Screening 2002 (NRS2002) tool [17].

This study is a retrospective analysis of data collected from the electronic medical record system of the hospital. The study protocol was submitted to and approved by the Institutional Review Board of Sun Yat-sen University Cancer Center (IRB of SYSUCC) (No. B2025-052-01). In accordance with the decision of the ethics committee, this study adheres to the principles of the Declaration of Helsinki and Good Clinical Practice guidelines. The IRB of SYSUCC waived the requirement for informed consent due to the retrospective nature of the study. All patient data were anonymized to ensure patient privacy.

Treatment and toxicities

All patients received conventional fractionated radiotherapy with 6–8 MeV X-rays, delivered using techniques including three-dimensional conformal radiotherapy (3D-CRT), intensity-modulated radiation therapy (IMRT), or volumetric modulated arc therapy (VMAT). Treatment planning was based on computed tomography (CT) scans with 3–5 mm slice, acquired using Monaco or Varian radiation systems. The gross tumor volume (GTV) included the esophageal tumor identified by CT and endoscopy, as well as enlarged regional lymph nodes on pre-treatment CT or PET-CT. The clinical target volume (CTV) comprised a 0.5–1.0 cm radial margin around the GTV and a 3 cm craniocaudal margin. The planning target volume (PTV) was defined as a 5–8 mm margin around the CTV. Adverse events were graded using the Common Terminology Criteria for Adverse Events v4.0 (CTCAE v4.0).

All patients received platinum- or fluoropyrimidine-based chemotherapy, administered intravenously (concurrently or induction) or orally (e.g. S-1 or capecitabine). Concurrent chemotherapy was delivered weekly during radiotherapy, while oral agents were administered daily. A subset of patients received 1–2 cycles of induction chemotherapy on three-week regimens prior to dCRT.

The specific anti-PD-1 agent, timing of administration relative to dCRT (induction, concurrent, and/or consolidation), and the number of cycles administered were recorded. Patients receiving ≥ 2 cycles of any anti-PD-1 antibody (nivolumab, pembrolizumab, toripalimab, sintilimab, tislelizumab, camrelizumab, or serplulimab) at standard three-week intervals—before, during, or after radiotherapy—were assigned to the immunotherapy (IO) group. Treatment continued until unacceptable toxicity or discontinuation at the treating physician’s discretion; although longer administration (up to 1–2 years) was permitted in routine practice, actual exposure in this elderly cohort was generally limited (median cycles reported in Results).

Statistical analysis

Clinical characteristics and adverse event rates were compared using Pearson’s chi-square test or two-tailed Fisher’s exact test. Survival outcomes were estimated using the Kaplan-Meier method and compared by the log-rank test. Progression-free survival (PFS) and overall survival (OS) were defined from the date of pathological diagnosis to disease progression or death, respectively. Prognostic factors were identified using univariate and multivariate Cox regression analyses. A hazard ratio (HR) < 1 indicated a reduced risk of progression or death in the IO group versus the dCRT group. Variables with p < 0.10 in univariate analysis were included in the multivariate models. Variables remaining statistically significant in multivariate analysis were considered independent prognostic factors.

To reduce potential confounding, propensity score matching (PSM) was performed at a 1:1 ratio with a caliper of 0.2, matching for gender, age, Karnofsky Performance Status (KPS), tumor length, tumor location, and TNM stage. After matching, standardized mean differences for all variables were < 10%. A two-sided p < 0.05 was considered statistically significant. All analyses were conducted using IBM SPSS Version 29.0 or R Version 4.4.1.

Results

Patients’ clinical characteristics

Between March 2009 and November 2023, 1340 patients diagnosed with ESCC and treated with dCRT (with or without immunotherapy) were screened, and 241 eligible patients were enrolled (Figure S1). Of these, 130 (53.9%) received dCRT alone (dCRT group) between March 2009 and November 2023, while 111 (46.1%) received dCRT combined with immunotherapy (IO group) between June 2019 and November 2023.

Baseline characteristics are summarized in Table 1. The IO group had higher proportions of males (74.8% vs 56.9%, p = 0.004), patients aged 70–75 years (76.6% vs 60%, p = 0.006), smokers (64.0% vs 46.2%, p = 0.006), and those receiving 50–60 Gy radiotherapy (84.7% vs 72.3%, p = 0.021). Comorbidities were present in 48.6% (120/241), including diabetes, hypertension, heart disease, chronic pulmonary disease, and liver disease. CCI distribution was comparable between groups (p = 0.655).

Table 1.

Baseline characteristics of the patients in the entire and matched cohorts.

Characteristic	Entire cohort before PSM (n = 241)		p value	Matched cohort after PSM (n = 220)		p value
	dCRT group	IO group		dCRT group	IO group
	(n = 130)%	(n = 111)%		(n = 110)%	(n = 110)%
Demographic characteristics
Sex
Male	74 (56.9%)	83 (74.8%)	0.004*	70 (63.6%)	82 (74.5%)	0.08
Female	56 (43.1%)	28 (25.2%)		40 (36.4%)	28 (25.5%)
Age, years
Median (IQR)	74 (71–77)	72 (71–74)		73 (71–77)	72 (71–74)
<75	78 (60.0%)	85 (76.6%)	0.006*	71 (64.5%)	84 (76.4%)	0.055
≥75	52 (40.0%)	26 (23.4%)		39 (35.5%)	26 (23.6%)
Smoking history
No	70 (53.8%)	40 (36.0%)	0.006*	53 (48.2%)	40 (36.4%)	0.076
Yes	60 (46.2%)	71 (64.0%)		57 (51.8%)	70 (63.6%)
Performance status and comorbidity
KPS
90–100	89 (68.5%)	85 (76.6%)	0.161	79 (71.8%)	84 (76.4%)	0.442
70–80	41 (31.5%)	26 (23.4%)		31 (28.2%)	26 (23.6%)
CCI
5	63 (48.5%)	57 (51.4%)	0.655	53 (48.2%)	56 (50.9%)	0.686
≥6	67 (51.5%)	54 (48.6%)		57 (51.8%)	54 (49.1%)
Tumor factors
Tumor length, cm
Median (IQR)	5 (3.9–6.2)	5.0 (4.0–6.6)		5.0 (4.0–6.2)	5.0 (4.0–6.6)
≤5cm	72 (55.4%)	58 (52.3%)	0.627	59 (53.6%)	58 (52.7%)	0.893
>5cm	58 (44.6%)	53 (47.7%)		51 (46.4%)	52 (47.3%)
Tumor location
Proximal	41 (31.5%)	37 (33.3%)	0.767	34 (30.9%)	36 (32.7%)	0.772
Middle thoracic &  Lower thoracic	89 (68.5%)	74 (66.7%)		76 (69.1%)	74 (67.3%)
T stage
T1–2	33 (25.4%)	19 (17.1%)	0.12	29 (26.4%)	19 (17.3%)	0.103
T3–4	97 (74.6%)	92 (82.9%)		81 (73.6%)	91 (82.7%)
N stage
N0–1	46 (35.4%)	39 (35.1%)	0.968	39 (35.5%)	39 (35.5%)	1
N2–3	84 (64.6%)	72 (64.9%)		71 (64.5%)	71 (64.5%)
TNM stage
I–III	70 (53.8%)	55 (49.5%)	0.506	60 (54.5%)	55 (50.0%)	0.5
IV	60 (46.2%)	56 (50.5%)		50 (45.5%)	55 (50.0%)
Treatment factors
Radiation dose, Gy
Median (IQR)	55.9 (50.4–60.0)	50.4 (50.0–56.0)		54.0 (50.4–59.94)	50.4 (50.0–55.92)
50–60	94 (72.3%)	94 (84.7%)	0.021*	83 (75.5%)	93 (84.5%)	0.092
≥60	36 (27.7%)	17 (15.3%)		27 (24.5%)	17 (15.5%)
Concurrent chemotherapy
without	15 (11.5%)	15 (13.5%)	0.643	9 (8.2%)	15 (13.6%)	0.194
with	115 (88.5%)	96 (86.5%)		101 (91.8%)	95 (86.4%)
Laboratory & nutrition
Baseline albumin, g/L
Median (IQR)	42.0 (40.0–44.2)	42.3 (40.1–44.8)		41.9 (40.5–44.1)	42.4 (40.1–44.8)
<40	48 (36.9%)	41 (36.9%)	0.626	40 (36.4%)	39 (35.5%)	0.579
≥40	82 (63.1%)	70 (63.1%)		70 (63.6%)	71 (64.5%)
NRS2002
2	51 (39.2%)	56 (50.5%)	0.081	46 (41.8%)	56 (50.9%)	0.176
≥3	79 (60.8%)	55 (49.5%)		64 (58.2%)	54 (49.1%)

PSM: Propensity Score Matching; KPS: Karnofsky Performance Status; NRS2002: Nutritional Risk Screening 2002; CCI: Charlson Comorbidity Index.

^*Statistically significant.

To reduce selection bias and confounding, 1:1 PSM was performed based on gender, age, KPS, tumor length, tumor location, and TNM stage, yielding 220 patients (110 per group) with balanced baseline characteristics (all p > 0.05). Post-PSM, the IO and dCRT groups were comparable in male predominance (74.5% vs 63.6%), median age (72 vs 73 years), and comorbidity burden (CCI = 5: 50.9% vs 48.2%; CCI ≥ 6: 49.1% vs 51.8%). Tumor characteristics, including T3–4 stage (82.7% vs 73.6%), N2-3 involvement (64.5% each), and stage IV distribution (50.0% vs 45.5%), were similar. Treatment parameters, such as 50–60 Gy radiotherapy (84.5% vs 75.5%) and concurrent chemotherapy (86.4% vs 91.8%), were also comparable.

Treatment characteristics of the matched cohort are detailed in Table S1. To assess the potential impact of epoch bias on staging, we analyzed the utilization of PET-CT. Throughout the study period, the use of PET-CT for initial staging remained consistently low and was comparable between groups both before and after PSM (Before PSM: dCRT 13.1% vs IO 9.0%, p = 0.318; After PSM: dCRT 14.5% vs IO 9.1%, p = 0.210). With respect to radiotherapy, the vast majority of patients in both groups were treated with modern techniques: 91.8% (101/110) in the dCRT group and 100% (110/110) in the IO group received IMRT or VMAT. Only a small proportion of patients in the dCRT group (8.2%) were treated with 3D-CRT. Regarding systemic therapy, the distribution of specific chemotherapy regimen types (e.g. taxane/platinum, fluorouracil/platinum) among treated patients was similar between groups during both the induction (p = 0.543) and concurrent phases (p = 0.087). However, patients in the IO group more frequently received induction chemotherapy (89.1% vs 41.8%) and a greater median number of total chemotherapy cycles (4 vs 2), reflecting evolving practice patterns in the immunotherapy era. In the IO group, six different anti-PD-1 agents were used. The three most common were toripalimab (30.0%), camrelizumab (23.6%), and sintilimab (19.1%), collectively accounting for 72.7% of patients receiving immunotherapy. Immunotherapy administration was heterogeneous: the majority of patients received anti-PD-1 therapy as induction prior to radiotherapy, with smaller proportions receiving it concurrently with radiotherapy and/or as consolidation afterward. The median number of anti-PD-1 cycles administered was 3 (IQR 2–6).

Survival outcomes

Survival outcomes were compared in unmatched and matched cohorts. Before PSM, median follow-up was 30.7 months (IQR 27.8–33.6) for the IO group and 40.2 months (IQR 33.7–46.6) for the dCRT group (Table S2). In the unmatched cohort, OS (HR = 0.85, 95%CI: 0.58–1.24, p = 0.410) and PFS (HR = 0.78, 95%CI: 0.55–1.10, p = 0.154) were not significantly different (Figure 1A,B). In the matched cohort, the median follow-up was 30.7 months (IQR 27.8–33.6) for the IO group and 38.6 months (IQR 25.7–51.6) for the dCRT group, reflecting the later clinical adoption of immunotherapy (Table S2). No significant differences were observed in OS (HR = 0.86, 95%CI: 0.58–1.28, p = 0.467) or PFS (HR = 0.79, 95%CI: 0.55–1.13, p = 0.194) (Figure 1C,D). Numerically, IO patients had longer median PFS in both unmatched (29.8 vs 18.6 months) and matched cohorts (29.8 vs 17.8 months), but differences were not statistically significant (Table S2).

Figure 1.

Kaplan–Meier curves for OS and PFS. (A) OS in the overall cohort of two groups. (B) PFS in the overall cohort of two groups. (C) OS in the propensity score-matched cohort of two groups. (D) PFS in the propensity score-matched cohort of two groups.

A sensitivity analysis restricted to the three most commonly used anti-PD-1 agents (toripalimab, camrelizumab, and sintilimab) yielded results consistent with the primary analysis for both PFS (HR = 0.77, 95% CI: 0.52–1.15; p = 0.206) and OS (HR = 0.87, 95% CI: 0.57–1.34; p = 0.536) (Table S3). Formal testing for heterogeneity of treatment effect across different agents within the IO group did not show significant differences for PFS (P for interaction = 0.596) or OS (P for interaction = 0.661), although these analyses are limited by sample size.

To explore potential factors underlying the dissociation between PFS and OS, we analyzed post-progression treatment patterns among matched patients with documented disease progression (Table S4). Among these patients, subsequent use of anti-PD-1 therapy was numerically more frequent in the IO group than in the dCRT group (35.2% vs 20.6%).

Risk factors of OS and PFS

Univariate and multivariate Cox analyses evaluated clinicopathological and treatment-related variables. Univariate analysis identified advanced TNM stage, high nutritional risk (NRS2002 ≥ 3), and high CCI (≥ 6) as predictors of shorter OS and PFS. Multivariate analysis confirmed high nutritional risk as an independent predictor of worse OS (HR = 1.69, 95%CI: 1.16–2.48, p = 0.007), while advanced TNM stage (HR = 1.55, 95%CI: 1.08–2.21, p = 0.017) and high nutritional risk (HR = 1.62, 95%CI: 1.14–2.30, p = 0.007) independently predicted shorter PFS (Table 2 and Table S5).

Table 2.

Univariate and multivariate analyses for overall survival in the entire and matched cohorts.

Variable	Entire cohort before PSM (n = 241)				Matched cohort after PSM (n = 220)
	Univariate		Multivariate		Univariate		Multivariate
	Hazard ratio (95%CI)	p value	Hazard ratio (95%CI)	p value	Hazard ratio (95%CI)	p value	Hazard ratio (95%CI)	p value
Sex (female vs. male)	0.83 (0.55–1.23)	0.345			0.69 (0.44–1.09)	0.111
Age (≥75 vs. <75 years)	1.02 (0.69–1.51)	0.913			1.13 (0.74–1.72)	0.580
Smoking history (yes vs. no)	1.01 (0.70–1.46)	0.949			1.07 (0.72–1.59)	0.740
Alcoholic history (yes vs. no)	0.90 (0.60–1.35)	0.627			0.91 (0.60–1.38)	0.646
Tumor length (>5 vs. ≤5 cm)	1.19 (0.83–1.71)	0.351			1.10 (0.74–1.62)	0.639
Tumor location (middle & lower vs. proximal)	1.17 (0.79–1.74)	0.436			1.29 (0.84–1.98)	0.250
TNM stage (IV vs. I–III)	1.58 (1.09–2.28)	0.016*	1.47 (0.99–2.16)	0.054	1.61 (1.09–2.38)	0.017*	1.48 (0.98–2.25)	0.060
Concurrent chemotherapy (yes vs. no)	0.92 (0.54–1.56)	0.747			1.04 (0.56–1.95)	0.898
Radiation dose (≥60 vs. 50–60 Gy)	1.21 (0.79–1.84)	0.382			1.05 (0.65–1.70)	0.831
Immunotherapy (yes vs. no)	0.85 (0.58–1.24)	0.410			0.86 (0.58–1.28)	0.467
KPS (70–80 vs. 90–100)	0.91 (0.60–1.38)	0.664			1.00 (0.64–1.57)	0.998
Baseline albumin (≥40 vs. <40)	0.99 (0.68–1.45)	0.971			0.93 (0.63–1.38)	0.721
NRS2002 (≥3 vs. 2)	1.68 (1.15–2.46)	0.007*	1.69 (1.16–2.48)	0.007*	1.62 (1.09–2.41)	0.017*	1.64 (1.11–2.44)	0.014*
CCI (≥6 vs. 5)	1.41 (0.98–2.03)	0.068	1.26 (0.86–1.86)	0.242	1.55 (1.04–2.29)	0.030*	1.35 (0.89–2.05)	0.153

Variables with a p value less than 0.10 in the univariate analysis were included in the multivariate model.

PSM: Propensity Score Matching; KPS: Karnofsky Performance Status; NRS2002: Nutritional Risk Screening 2002; CCI: Charlson Comorbidity Index; CI: confidence interval.

^*Statistically significant.

Post-PSM, findings were consistent. NRS2002 ≥ 3 remained independently associated with worse OS (HR = 1.64, 95%CI: 1.11–2.44, p = 0.014), and both advanced TNM stage (HR = 1.52, 95%CI: 1.04–2.22, p = 0.030) and NRS2002 ≥ 3 (HR = 1.57, 95%CI: 1.09–2.26, p = 0.016) independently predicted shorter PFS. Immunotherapy was not significantly associated with OS or PFS in either cohort (Table 2 and Table S5).

Subgroup analyses

Exploratory subgroup analyses were performed based on KPS, baseline albumin, and CCI to identify potential signals. In the unmatched cohort, the direction of subgroup effects was generally consistent with the matched cohort, with only minor differences in hazard ratio estimates and confidence intervals (Figure S2). Therefore, the following results focus on the matched cohort.

Specifically, in the matched cohort, patients with KPS 90–100 had significantly improved PFS in the IO group compared to the dCRT group. The median PFS was not reached in the IO group versus 17.2 months in the dCRT group. The 12-, 24-, and 36-month PFS rates were 72.5%, 56.8%, and 50.2% in the IO group, compared with 66.8%, 44.0%, and 35.2% in the dCRT group (p = 0.043) (Figure 2A). OS differences were not significant (median 37.1 vs 29.0 months; p = 0.113) (Figure 2B). Patients with KPS 70–80 showed no PFS or OS benefit (Figure S3A,B).

Figure 2.

Kaplan–Meier curves for PFS and OS in the propensity score-matched cohort by subgroup. (A) PFS for patients in two groups with KPS 90–100. (B) OS for patients in two groups with KPS 90–100. (C) PFS for patients in two groups with CCI = 5. (D) OS for patients in two groups with CCI = 5. (E) PFS for patients in two groups with baseline albumin ≥ 40 g/L. (F) OS for patients in two groups with baseline albumin ≥ 40 g/L.

Given the correlation between general physical condition and nutritional status, we further examined patients’ CCI and baseline albumin levels. Patients with CCI = 5 showed improved PFS in the IO group compared to the dCRT group (HR = 0.54, 95%CI: 0.31–0.96, p = 0.035) (Figure 2C). The OS for these patients was not significantly different between groups (HR = 0.59, 95% CI 0.32–1.09, p = 0.092) (Figure 2D). In contrast, no survival benefit was observed in those with CCI ≥ 6 (Figure S3C,D). When stratified by baseline albumin, patients with albumin ≥ 40 g/L had improved PFS with immunotherapy (HR = 0.57, 95%CI: 0.36–0.92, p = 0.022) (Figure 2E), whereas their OS was not significantly different (HR = 0.63, 95%CI: 0.38–1.05, p = 0.075) (Figure 2F). No significant differences in PFS or OS were found among patients with albumin <40 g/L (Figure S3E,F).

Interaction analyses in the overall cohort indicated significant modification effects on OS for KPS (p = 0.034) and CCI (p = 0.036), and on PFS for KPS (p = 0.028), baseline albumin (p = 0.040), and CCI (p = 0.032) (Figure S4). Post-PSM, baseline albumin approached significance for PFS (p = 0.050), while no significant interactions were observed for OS (Figure 3).

Figure 3.

Forest plots of hazard ratios and 95% confidence intervals for subgroup analyses in the propensity score-matched cohort. (A) Forest plot for OS in two groups. (B) Forest plot for PFS in two groups.

In these exploratory analyses, the estimated treatment effect on PFS was generally consistent across most baseline characteristics. Nominal interaction signals were observed in a small number of subgroups (e.g. KPS 90–100, albumin ≥40 g/L, and CCI = 5). However, it should be noted that these subgroups had limited sample sizes, which reduces statistical power and increases the risk of false-positive findings. Therefore, these results should be interpreted as hypothesis-generating.

Safety profile

Toxicities in the matched cohort were comparable between groups (Table 3). Anemia was most common (IO group: 81.8% vs dCRT group: 87.3%, p = 0.351). Leukopenia (63.6% vs 55.5%, p = 0.272) and neutropenia (38.2% vs 33.6%, p = 0.574) were slightly higher in the IO group, but not significant. Pneumonia occurred in 31.2% (29/93) of the IO group and 33.0% (35/106) of the dCRT group patients (p = 0.901), with grade 3–4 events observed in 3.2% (3/93) versus 1.9% (2/106) (p = 0.666). Rates of esophagitis (any grade: 65.5% vs 63.6%; grade 3–4: 5.5% vs 2.7%) and transaminitis (23.6% vs 26.4%) were similar. Hypothyroidism occurred in 21.8% (24/110) of the IO group; because routine thyroid function monitoring was not systematically performed in the dCRT group, a direct comparison of incidence is not appropriate. For pulmonary events, while the overall rates of pneumonia were comparable (IO: 31.2% vs dCRT: 33.0%), the retrospective design limits reliable etiologic distinction between radiation pneumonitis, immune-related pneumonitis, and infectious pneumonia. Overall, the addition of immunotherapy did not significantly increase the incidence of reported toxicities.

Table 3.

Toxicities in the 220 patients of the matched cohort.

Toxicities	dCRT group		IO group		p value
	count (n = 110)	percentage (%)	count (n = 110)	percentage (%)
Anemia	96	87.3	90	81.8	0.351
Grade 1–2	89	80.9	79	71.8	0.153
Grade 3–4	7	6.4	11	10.0	0.461
Leucopenia	61	55.5	70	63.6	0.272
Grade 1–2	47	42.7	60	54.6	0.106
Grade 3–4	14	12.7	10	9.1	0.517
Neutropenia	37	33.6	42	38.2	0.574
Grade 1–2	23	20.9	34	30.9	0.124
Grade 3–4	14	12.7	8	7.3	0.261
Thrombopenia	38	34.6	44	40.0	0.486
Grade 1–2	32	29.1	36	32.7	0.662
Grade 3–4	6	5.5	8	7.3	0.782
Lymphopenia	106	96.4	108	98.2	0.683
Grade 1–2	20	18.2	24	21.8	0.613
Grade 3–4	86	78.2	84	76.4	0.872
Pericardial effusiona	22	22.0	23	26.4	0.592
Pleural effusiona	28	28.3	27	31.0	0.803
Pneumoniaa	35	33.0	29	31.2	0.901
Grade 1	25	23.6	19	20.4	0.716
Grade 2	8	7.6	7	7.5	1
Grade 3–4	2	1.9	3	3.2	0.666
Esophagitis	70	63.6	72	65.5	0.888
Grade 1–2	67	60.9	66	60.0	1
Grade 3–4	3	2.7	6	5.5	0.499
Transaminitis	29	26.4	26	23.6	0.756
Hyperthyroidisma	/	/	0	0	/
Hypothyroidisma	/	/	24	21.8	/
Grade 1–2	/	/	24	21.8	/
Grade 3–4	/	/	0	0	/

^a Percentages calculated based on available cases; not all patients completed every assessment.

/: Data not routinely collected.

Discussion

Interpretation of the study findings in the context of existing evidence

The addition of immunotherapy to chemoradiotherapy has been explored in phase II clinical trials for locally advanced ESCC, showing higher clinical complete response (cCR) and pathological complete response (pCR) rates compared to previous standards [15]. However, no phase III trial has demonstrated its superiority over chemoradiotherapy alone. Prospective data in elderly patients are especially limited. Therefore, this study aimed to evaluate real-world outcomes of dCRT combined with immunotherapy in contemporary supportive care frameworks, comparing them with historical outcomes achieved with dCRT alone in elderly ESCC patients.

As this is a retrospective analysis spanning different treatment eras, it is important to note that the independent contribution of immunotherapy cannot be fully separated from other contemporaneous improvements in medical care. Thus, our findings should be considered exploratory and hypothesis-generating. They provide valuable context for clinical decision-making in elderly ESCC patients while emphasizing the need for prospective, era-matched trials to better define the incremental benefit of adding immunotherapy to dCRT.

Progression-free survival signal without overall survival benefit

In this cohort of 220 elderly patients (≥70 years) with locally advanced ESCC, the addition of anti-PD-1 antibodies to dCRT did not result in significant improvements in PFS or OS in either the overall cohort or after PSM. However, a numerical trend toward longer PFS was observed. The dissociation between the observed trend in PFS and the absence of an OS benefit is likely multifactorial.

Patients in the IO group received more intensive systemic therapy (e.g. higher induction chemotherapy use), reflecting evolving treatment practices. To assess whether this difference could explain the PFS trend, we note that a prior prospective phase II trial in unresectable ESCC showed that adding induction chemotherapy to dCRT did not significantly improve survival compared with dCRT alone [18,19]. Additionally, our categorization of chemotherapy as ‘platinum- or fluoropyrimidine-based’ aligns with prospective randomized trials in locally advanced ESCC, which demonstrated comparable survival outcomes with various paclitaxel-based combinations in definitive chemoradiation [20]. These findings suggest that differences in chemotherapy intensity and regimen type are unlikely to fully explain the outcomes observed in our cohort.

We also examined whether post-progression therapies could account for the observed dissociation. Analysis of available data showed no significant difference in the use of salvage immunotherapy between groups (Table S4), making differential subsequent therapy an unlikely explanation for the results. Thus, the observed PFS trend likely reflects the effects of a contemporary treatment strategy rather than a definitive causal effect of immunotherapy. Notably, the relatively limited immunotherapy exposure in this elderly cohort (median of 3 cycles) suggests that prolonged maintenance therapy is unlikely to account for the observed PFS signal. However, given the inherent limitations of the retrospective design and the incomplete availability of second-line treatment data, we cannot exclude the possibility that unrecorded or heterogeneous salvage therapies may have influenced OS outcomes. These findings should be interpreted with appropriate caution.

The inclusion of multiple PD-1 inhibitors reflects real-world clinical practice. We note that phase III trials in advanced ESCC have reported consistent efficacy across different PD-1 inhibitors when combined with chemotherapy [10,11,13,14,21]. Additionally, prospective head-to-head comparisons in other cancers support the biological plausibility that the observed outcomes in our cohort reflect a class effect of PD-1 blockade rather than being driven by a single specific agent [22].

Potential biological explanations: immunosenescence and competing risks

In elderly patients, age-related biological factors may attenuate the benefits of immunotherapy. Immunosenescence involves the accumulation of terminally differentiated, hypoproliferative CD8⁺ T cells that lack the CD28 co-stimulatory molecule and often express CD57 and PD-1 [23,24]. This deficit impairs the efficacy of anti-PD-1 therapy, as CD28 signaling is crucial for the clonal expansion of PD-1⁺ T cells following checkpoint blockade [25]. Additionally, the chronic, low-grade inflammation of ‘inflammaging’ (e.g. elevated IL-6) can suppress T-cell activation within tumors [25], potentially obscuring survival benefits.

Competing risks of non-cancer mortality, such as cardiovascular or respiratory diseases, are more common in elderly populations and may further obscure potential survival gains [26]. Moreover, functional decline following disease progression can limit the tolerance for subsequent therapies [27], thereby attenuating OS benefits despite delayed progression.

Although our dCRT group’s 3-year OS (46.8%) and PFS (38.5%) were consistent with historical rates of 43.4%–46.2% and ∼37.3% [8,9], these collective factors underscore the complexity of translating PFS gains into OS benefits for elderly patients with ESCC.

Nutritional status as a key host-related determinant of outcomes

A key finding of this study is the strong prognostic impact of nutritional status. High nutritional risk (NRS2002 ≥ 3) was the only host-related factor independently associated with worse OS (HR = 1.64, 95%CI: 1.11–2.44, p = 0.014) and PFS (HR = 1.57, 95%CI: 1.09–2.26, p = 0.016), whereas performance status, baseline albumin, and comorbidity burden were not independent predictors in multivariate models. These results suggest that nutritional status may more directly reflect physiological reserve and treatment tolerance in elderly ESCC patients.

Biologically, nutrition is closely linked to immune competence. For instance, albumin stabilizes immunoglobulin G (IgG) [28], which may partly explain the association between hypoalbuminemia and poor outcomes. Malnutrition can exacerbate metabolic competition in the tumor microenvironment [29], potentially starving tumor-infiltrating lymphocytes and impairing anti-tumor immunity. Additionally, gut dysbiosis can impair PD-1 inhibitor efficacy [30,31], suggesting that modulating the microbiome could be a strategy to enhance treatment response. Furthermore, malnutrition-induced proinflammatory cytokines (e.g. TNF-α, IL-6, IL-1β) may upregulate immune checkpoint pathways [32,33], creating an immunosuppressive environment that could blunt the effects of immunotherapy. These mechanisms support the concept that nutritional status is not merely a passive prognostic marker but a modifiable host factor that can potentially enhance treatment outcomes. Targeted nutritional interventions, including optimization of protein intake and management of sarcopenia, may improve treatment tolerance and immune responsiveness [34,35], although this hypothesis requires validation in prospective studies.

These findings support the incorporation of routine nutritional screening, such as NRS2002, into pretreatment evaluation of elderly ESCC patients, with early supportive care involvement aimed at optimizing physiological reserve within a geriatric oncology framework.

Beyond its prognostic implications and potential impact on efficacy, the relationship between nutritional status and immune-related adverse events (irAEs) merits further exploration. Malnutrition is linked to gut barrier dysfunction and chronic, low-grade systemic inflammation, which may influence both the efficacy of immunotherapy and the host’s susceptibility to inflammatory toxicities [36]. For example, a compromised gut barrier could facilitate microbial translocation, potentially altering immune activation and toxicity profiles [37]. However, direct clinical evidence connecting pre-treatment nutritional status to irAE risk in the context of combined chemoradiotherapy and immunotherapy for ESCC is limited. As such, this observation should be viewed as a basis for future mechanistic and clinical studies, rather than a definitive conclusion.

Exploratory subgroup signals and the concept of the ‘fit elderly’

Exploratory subgroup analyses suggested that baseline physiological reserve may influence treatment efficacy. Patients with superior functional status (KPS 90–100), higher serum albumin levels (≥ 40 g/L), and lower comorbidity burden (CCI = 5) experienced greater PFS benefits from combination therapy. However, these subgroups were small, and the interaction P values were borderline, increasing the risk of false-positive findings. Accordingly, these subgroup signals are presented as preliminary observations that require confirmation in larger, prospective studies.

These readily measurable and reproducible markers effectively identified a physiologically fitter geriatric subgroup within our cohort. The absence of PFS improvement in patients with poorer markers suggests baseline immune competence, rather than tumor immunogenicity alone, may be a key determinant of PD-1 inhibitor efficacy. Although PFS gains did not translate into statistically significant OS benefits, the observed trend remains clinically informative.

Collectively, these exploratory findings suggest that a combination of KPS 90–100, serum albumin ≥40 g/L, and CCI <6 could serve as practical proxies for physiological fitness in elderly ESCC patients. KPS reflects functional reserve, CCI captures competing mortality risk, and albumin may indicate a less inflammatory and more immunologically competent state. Importantly, these factors are at least partially modifiable through nutritional support, rehabilitation, and comorbidity management. Incorporating these markers into a comprehensive geriatric assessment [38] may facilitate personalized, risk-adapted treatment strategies.

Taken together, these findings underscore the importance of assessing physiological fitness rather than chronological age in elderly ESCC. Using simple, readily available metrics to stratify patients into ‘fit’ or ‘vulnerable’ categories aligns with geriatric risk-adapted care and can help inform individualized decisions regarding intensive versus supportive treatment strategies.

Safety considerations in combined chemoradiotherapy and immunotherapy

Treatment-related toxicities were generally comparable between groups, suggesting that the addition of anti-PD-1 therapy did not substantially increase overall toxicity in appropriately selected elderly patients. Pneumonia, a clinically relevant adverse event, occurred at similar rates in both groups (IO group 31.2% vs dCRT group 33.0%; p = 0.901), including grade 3–4 events (3.2% vs 1.9%; p = 0.666), supporting pulmonary safety in this setting.

However, safety interpretation requires caution. Differences in hypothyroidism reporting primarily reflect surveillance bias, as routine thyroid monitoring was not performed in the dCRT group. In addition, the retrospective design precludes reliable etiologic distinction between radiation pneumonitis, immune-related pneumonitis, and infectious pneumonia, and limits the exclusion of pre-existing pulmonary conditions that may further confound attribution of pulmonary events. Longer-term prospective studies with standardized monitoring are needed to better characterize immune-related toxicities in elderly patients receiving combined-modality therapy.

Taken together, our findings have implications for clinical practice and future research. These results suggest that in clinical practice, treatment decisions for elderly ESCC patients should prioritize a comprehensive assessment of physiological fitness, with particular attention to nutritional status and functional capacity, rather than relying on chronological age alone or adopting a uniform strategy of treatment intensification. Future studies should prospectively validate these markers of fitness and explore interventions to optimize physiological reserve in this vulnerable population.

Limitations and strengths

This study has several important limitations that merit consideration. First and most importantly, epoch bias represents the principal limitation of this analysis. This retrospective study spans a 14-year period, during which secular improvements in multidisciplinary supportive care (e.g. antiemetics, nutritional support, and toxicity management), as well as evolving radiotherapy techniques and chemotherapy practices, may have conferred a nonspecific survival advantage to patients treated in the more recent era, including those in the IO group. Although detailed reporting suggests broad comparability in radiotherapy techniques and chemotherapy regimen types between groups, the influence of evolving standards of care cannot be fully excluded. Accordingly, our findings should be interpreted as reflecting real-world outcomes within a contemporary treatment context rather than as definitive evidence isolating the independent effect of immunotherapy.

Second, this study compares two fundamentally different treatment strategies: a contemporary paradigm incorporating immunotherapy and a historical dCRT-only approach. The absence of consolidation or maintenance immunotherapy in the dCRT group represents a major difference in treatment intensity, and immunotherapy timing was heterogeneous (induction, concurrent, and/or consolidation), limiting causal attribution and precluding determination of optimal sequencing.

Third, follow-up duration was shorter in the IO group, which may limit the maturity of overall survival estimates and the assessment of very late-onset immune-related adverse events, underscoring the need for continued follow-up. In addition, detailed chemotherapy dose-level data were not consistently available, necessitating the use of regimen type and cycle number as surrogate measures of systemic therapy intensity.

Finally, several limitations inherent to retrospective observational studies remain unavoidable. These include residual confounding from unmeasured factors (such as physician selection bias and the lack of standardized geriatric frailty assessments), limited statistical power for exploratory subgroup analyses, incomplete information on post-progression therapies, potential heterogeneity related to the use of multiple PD-1 inhibitors, and the absence of biomarker data (e.g. PD-L1 expression), which together restrict causal inference and mechanistic interpretation.

Despite these limitations, this study has several important strengths. It addresses a key evidence gap by focusing on elderly patients (≥70 years) with locally advanced ESCC, an underrepresented group in clinical trials, and provides valuable real-world data on the efficacy and safety of immunotherapy in this setting. Moreover, by integrating host-related factors such as nutritional status, functional performance, and comorbidity burden, this study highlights the central role of physiological reserve in treatment outcomes and supports the incorporation of geriatric assessment into immunotherapy decision-making.

Conclusion

In elderly patients with ESCC, adding anti–PD-1 antibodies to definitive chemoradiotherapy did not significantly improve OS or PFS but was associated with acceptable safety in a real-world setting. Exploratory findings indicate a potential PFS signal in selected physiologically fit patients, highlighting the relevance of physiological fitness in treatment decision-making and the need for prospective validation.

Funding Statement

This work was supported by grants from the Basic and Applied Basic Research Foundation of Guangdong Province (2025A1515010951).

Ethical approval

This study was ethically approved by the Institutional Review Board of Sun Yat-sen University Cancer Center (No. B2025-052-01).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Declaration of generative AI in scientific writing

During the preparation of this manuscript, the authors used ChatGPT-4o and Grammarly solely for grammar checking. After using these tools, all authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Data availability statement

The datasets generated and/or analyzed in this study are available from the corresponding author Qiaoqiao Li upon reasonable request.