Abstract

Objectives:

To explore the optimal timing of chemoimmunotherapy (CIT) combined with radiotherapy (RT) for oligometastatic esophageal squamous cell carcinoma (ESCC) and analyze prognostic factors.

Methods:

In this multicenter, retrospective study, we reviewed 142 patients with oligometastatic ESCC who received first-line RT and CIT between 2019 and 2024. Based on the timing of RT initiation relative to CIT, patients were categorized into three groups: RT-first (n=27), Concurrent (RT within the first 4 cycles of CIT, n=87), and Sequential (RT after 4 cycles of CIT, n=28). Inverse probability of treatment weighting (IPTW) and multivariable Cox regression were utilized to mitigate baseline confounding.

Results:

Following IPTW adjustment, baseline characteristics were well-balanced. The Concurrent group demonstrated a significantly superior overall survival (OS) compared to the RT-first and Sequential groups (median OS: 22.7 vs. 21.0 vs. 18.1 months, respectively; P = 0.018). Multivariable analysis confirmed that the concurrent strategy was an independent protective factor for OS (HR = 0.595, 95% CI: 0.354–0.999, P = 0.049). Progression-free survival (PFS) and anatomical patterns of first failure did not differ significantly among the three cohorts. Furthermore, the incidence of Grade ≥ 3 treatment-related adverse events was comparable across all groups (P > 0.05).

Conclusions:

In patients with oligometastatic ESCC, concurrent integration of RT within the early cycles of first-line chemoimmunotherapy is significantly associated with prolonged overall survival, without increasing severe toxicity, compared to upfront or delayed RT. These findings provide compelling real-world evidence to guide optimal treatment sequencing, warranting further validation in prospective trials.

1 Introduction

According to the 2022 National Cancer Report, Esophageal Cancer (EC) remains a major public health burden in China, ranking as the seventh most common malignancy and the fifth leading cause of cancer-related mortality in China (1). Due to its insidious onset, the majority of patients are diagnosed at a locally advanced or metastatic stage. Even with guideline-recommended first-line therapies, the median overall survival (mOS) remains poor, typically ranging from 12.4 to 17.2 months (2). Oligometastatic EC is generally considered an intermediate state between localized and widely metastatic disease (3), with reported 3-year and 5-year actuarial survival rates of 59.5% and 51.7%, respectively (4). The currently accepted definition of oligometastatic esophageal squamous cell carcinoma (ESCC) is: ≤5 metastatic sites distributed across ≤3 organs, or metastasis to a single regional lymph node (5–7).

In recent years, the standard of care for advanced ESCC has fundamentally shifted to immune checkpoint inhibitors (ICIs) combined with chemotherapy (chemoimmunotherapy, CIT). However, systemic therapy alone rarely achieves durable disease control in macroscopic lesions, often leading to acquired resistance and disease progression. A growing body of clinical evidence suggests that integrating local therapy—particularly radiotherapy (RT)—into the CIT backbone can significantly improve outcomes for oligometastatic ESCC (8–10). Radiotherapy not only achieves local cytoreduction but also acts as an immunological adjuvant. Radiation-induced tumor cell death releases tumor neoantigens and pro-inflammatory cytokines, modifying the tumor microenvironment and triggering an “in situ vaccine” effect that is associated with the potentiation of the systemic efficacy of ICIs. Consequently, European clinical practice guidelines for oligometastatic esophageal cancer now endorse the incorporation of RT with CIT (11).

Despite this theoretical synergy, an extensive review of the current literature reveals a critical knowledge gap regarding the optimal temporal sequencing of RT and CIT. The timing of RT integration—whether upfront, concurrently with early cycles of CIT, or sequentially as consolidation—remains highly controversial. Biologically, upfront or extensive RT may induce severe radiation-induced lymphopenia, depleting the circulating effector T-cells required for ICI efficacy (12, 13). Conversely, delaying RT too long may miss the optimal window for immune priming and be linked to an increased risk of systemic dissemination. While landmark studies in other thoracic malignancies (e.g., the PACIFIC trial) (14) have underscored the profound impact of chemoradiotherapy sequencing on survival, specific evidence in oligometastatic ESCC is sparse and contradictory. Several phase II trials and small retrospective series have explored either concurrent or sequential approaches independently, including studies such as PALACE-1, Duan et al., and Li et al. (5, 9, 15) yet direct comparative analyses evaluating the exact clinical timing of RT initiation within a real-world first-line CIT setting are virtually non-existent.

While systemic chemoimmunotherapy has become the standard first-line treatment for metastatic ESCC, the integration of local radiotherapy remains an area of active investigation. Several recent retrospective series have provided evidence for the benefit of adding RT to systemic regimens. For instance, Amin et al. (2024) demonstrated that the addition of RT to immunotherapy and chemotherapy significantly improved overall survival in patients with stage IV esophageal cancer, supporting the survival benefit of combined-modality treatment in advanced disease (16). Similarly, an analysis of the National Cancer Database by Tasoudis et al. (2024) demonstrated the feasibility and survival benefit of incorporating immunotherapy into multimodal chemoradiation strategies, further supporting the clinical value of combined-modality treatment (17). Furthermore, Wu et al. (2023) confirmed that CIT combined with RT offers superior outcomes compared to CIT alone in real-world metastatic ESCC cohorts (18). Collectively, these retrospective and real-world studies consistently support the clinical value of integrating radiotherapy into systemic immunotherapy-based strategies; however, they primarily focused on the presence or absence of radiotherapy rather than the optimal timing of its delivery. Importantly, none of these studies evaluated the comparative efficacy of different radiotherapy sequencing strategies within the same cohort, leaving the clinically relevant question of optimal RT timing largely unanswered.

However, despite the accumulating evidence supporting RT integration, the optimal temporal sequencing of radiotherapy—specifically whether RT should be delivered before, during, or after systemic chemoimmunotherapy—remains a major unresolved clinical question. Most previous studies evaluated concurrent or sequential approaches independently, without performing direct head-to-head comparisons among multiple timing strategies within the same clinical cohort.

Therefore, this multicenter real-world study was designed to directly compare survival outcomes and failure patterns among three distinct radiotherapy sequencing strategies: RT-first, concurrent RT, and sequential RT. By leveraging inverse probability of treatment weighting (IPTW) to mitigate baseline confounders, this study aims to evaluate the optimal sequence for integrating RT with CIT in oligometastatic ESCC, thereby providing clinically relevant evidence to inform the optimal integration of radiotherapy into first-line chemoimmunotherapy for oligometastatic ESCC.

2 Materials and methods2.1 Study design and patient population

This was a multi-center, retrospective study. We included patients with oligometastatic esophageal squamous cell carcinoma (ESCC) who were treated between January 1, 2019, and October 30, 2024, at three tertiary hospitals in China: the Affiliated Hospital of North Sichuan Medical College, Nanchong Central Hospital, and Suining Central Hospital. The study cohort consisted of patients with any of the nine classes of oligometastatic disease as defined by the joint expert consensus of the European Society for Radiotherapy and Oncology (ESTRO) and the European Organization for Research and Treatment of Cancer (EORTC). A detailed flowchart of the patient selection process is provided in the STROBE diagram (Figure 1).

STROBE flowchart.

We initially screened the institutional databases to identify a preliminary cohort of patients diagnosed with ESCC. To establish the final study cohort, patients were strictly required to meet all of the following Inclusion Criteria (1): Histologically confirmed ESCC with clinical Stage IV disease (as defined by the American Joint Committee on Cancer [AJCC] staging system), confirmed by imaging such as MRI, CT, or PET-CT (2). Presence of oligometastatic disease, defined as a total of ≤5 metastatic lesions in ≤3 organs, or metastasis to a single regional lymph node (3). An Eastern Cooperative Oncology Group (ECOG) performance status score of 0–2 (4). For patients with primary (synchronous) oligometastatic ESCC: receipt of at least four cycles of chemoimmunotherapy, with radiotherapy administered to the primary esophageal tumor (IMRT, 50–60 Gy) and/or metastatic sites (IMRT/SBRT, 27–60 Gy) (5). For patients with metachronous oligometastatic ESCC: receipt of at least four cycles of chemoimmunotherapy, with radiotherapy administered to the metastatic sites (IMRT/SBRT, 27–60 Gy). Re-irradiation was permissible for local esophageal recurrence.

Subsequently, patients who met any of the following conditions were removed from the cohort based on the Exclusion Criteria (1): A confirmed diagnosis of a second primary malignancy within the previous five years (2). Incomplete clinical data or loss to follow-up (3). Failure to complete at least four cycles of chemoimmunotherapy or the full course of prescribed radiotherapy (4). Presence of severe comorbidities, such as uncontrolled hypertension, significant heart disease, diabetes mellitus, or chronic obstructive pulmonary disease.

Some patients with oligometastatic esophageal cancer present with significant symptoms from their metastatic lesions, for whom upfront radiotherapy may be administered for palliative relief. This study protocol was approved by the Institutional Ethics Review Committee (Approval No. 2025ER206-1). Due to the retrospective nature of the study, the requirement for individual patient informed consent was waived. All procedures were conducted in accordance with the ethical principles of the Declaration of Helsinki. Based on the timing and sequence of radiotherapy relative to chemoimmunotherapy, patients were categorized into three groups: the RT-first group (radiotherapy administered before the start of chemoimmunotherapy, n = 27), the Concurrent group (radiotherapy administered within the first four cycles of chemoimmunotherapy, n = 87), and the Sequential group (radiotherapy administered after the completion of four cycles of chemoimmunotherapy, n = 28).

2.2 Treatment protocols2.2.1 Systemic chemoimmunotherapy

Rather than a generalized approach, all 142 patients in this retrospective cohort received specific first-line systemic chemoimmunotherapy based on institutional standards and individual patient profiles. The immunotherapy backbone predominantly consisted of PD-1 inhibitors, specifically camrelizumab (n = 96), tislelizumab (n = 40), or sintilimab (n = 6), administered intravenously every 3 weeks. Concurrent chemotherapy regimens primarily comprised a platinum-based agent (cisplatin or nedaplatin) combined with either a taxane (paclitaxel) or a fluoropyrimidine (5-fluorouracil or capecitabine). Patients typically received 4 to 6 cycles of this combination therapy, followed by immune maintenance therapy until disease progression or unacceptable toxicity.

2.2.2 Radiotherapy regimens

For radiotherapy, involved-field radiotherapy (IFRT) was strictly utilized for the primary esophageal tumor and clinically involved regional lymph nodes, deliberately avoiding elective nodal irradiation (ENI). This precise targeting strategy was adopted to minimize the irradiation volume, thereby sparing circulating lymphocytes and correlating with a reduced risk of severe radiation-associated lymphopenia to support the efficacy of concurrent immune checkpoint inhibitors.

In our cohort, 75 patients received RT to the primary tumor using conventionally fractionated intensity-modulated radiation therapy (IMRT), with a median delivered dose of 50.4 Gy (interquartile range [IQR]: 50.4–54.0 Gy; range: 50.4–60.0 Gy). Meanwhile, 67 patients received RT targeting specific oligometastatic lesions. Modalities such as IMRT or stereotactic body radiation therapy (SBRT) were applied based on the specific anatomical site and tumor volume, with an overall median dose of 50.0 Gy (IQR: 44.0–50.0 Gy; range: 30.0–60.0 Gy). The dose-fractionation schedules for metastatic sites were highly individualized to achieve optimal local control while respecting normal tissue constraints: lung metastases (e.g., 60 Gy in 15 fractions), bone metastases (e.g., 20–30 Gy in 5–10 fractions), brain metastases (e.g., 30–52.5 Gy in 10–15 fractions), liver metastases (e.g., 48 Gy in 12 fractions), adrenal metastases (e.g., 30 Gy in 10 fractions), and lymph node metastases (e.g., 50–60 Gy in 25–30 fractions).

2.2.3 Definition of treatment sequencing

The primary intervention evaluated in this study was the temporal integration of local RT with systemic chemoimmunotherapy. Based on the exact clinical timing of RT initiation relative to the systemic therapy cycles, patients were categorized into three distinct strategies (1): RT-first group (n = 27): Radiotherapy was administered and completed prior to the initiation of chemoimmunotherapy, often necessitated by significant local symptoms requiring immediate palliative relief (2). Concurrent group (n = 87): Radiotherapy was seamlessly integrated within the first four cycles of induction chemoimmunotherapy (3). Sequential group (n = 28): Radiotherapy was administered as a consolidative local therapy only after the completion of at least four full cycles of chemoimmunotherapy.

2.3 Efficacy and toxicity assessment

Treatment efficacy was evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Responses were categorized as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD). The disease control rate (DCR) was calculated as the proportion of patients who achieved CR, PR, or SD: DCR = ([CR + PR + SD]/total number of patients) × 100%.

Progression-free survival (PFS) was defined as the time from the initiation of the first treatment for oligometastatic disease (either radiotherapy or chemoimmunotherapy) to the date of disease progression or the last follow-up. Overall survival (OS) was defined as the time from the diagnosis of oligometastatic disease to the date of death from any cause or the last follow-up. Adverse events were graded using the Common Terminology Criteria for Adverse Events (CTCAE), version 5.0.

2.4 Statistical analysis

All statistical analyses were performed using R software (version 4.5.1). Baseline patient characteristics were described and stratified by treatment sequencing group (RT-first, Concurrent, and Sequential). Categorical variables are presented as frequencies and percentages, while continuous variables are presented as medians and interquartile ranges (IQR). The Standardized Mean Difference (SMD) was calculated to quantify the extent of imbalance in baseline covariates, with an SMD > 0.2 indicating a potential imbalance.

To robustly evaluate the association between treatment sequencing and survival while mitigating “confounding by indication,” we employed a causal inference framework utilizing Stabilized Inverse Probability of Treatment Weighting (IPTW). First, a Directed Acyclic Graph (DAG) was constructed based on existing clinical knowledge to formalize prior structural assumptions. By applying the back-door criterion, a minimal sufficient adjustment set was identified, comprising ECOG performance status, type of oligometastasis, number of metastatic lesions, and baseline neutrophil-to-lymphocyte ratio (NLR).

Rather than relying solely on traditional multivariable Cox regression—which is constrained by assumptions of linear additivity and is prone to overfitting when adjusting for numerous confounders relative to the event size—we utilized a dual-model propensity score strategy to create a balanced pseudo-population. Alongside a conventional multinomial logistic regression (LR) model, we developed a gradient boosting machine (GBM) model. As a non-parametric machine learning algorithm, GBM automatically captures complex non-linear relationships and multi-way interactions among the covariates without requiring manual specification, thereby achieving superior covariate balance. To prevent model overfitting and avoid the generation of extreme weights, the GBM tuning process was rigorously optimized using 5-fold cross-validation to determine the optimal number of decision trees.

Stabilized IPTW weights were then calculated. The model demonstrating the optimal covariate balance (defined as yielding the lowest SMDs among core confounders, targeting SMD < 0.2) was selected for all subsequent outcome analyses. This IPTW pipeline effectively separates the confounding adjustment (design phase) from the outcome evaluation (analysis phase).

The primary outcome analysis was conducted on the IPTW-weighted cohort. Weighted Kaplan-Meier curves were plotted to visualize survival distributions, and the weighted log-rank test was used to assess overall differences. To evaluate the independent clinical association of treatment timing with survival, a weighted multivariable Cox proportional hazards model was constructed to calculate adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs). To adjust for potential residual confounding post-weighting, the specific radiotherapy site (primary vs. metastatic) was included as an additional covariate in this final model. Pairwise comparisons among the three groups were performed, and the Bonferroni correction was applied to control the overall Type I error rate, with a two-sided P < 0.0167 (0.05/3) considered statistically significant for these specific comparisons.

Finally, to explore independent prognostic factors, the treatment groups and other covariates were included in a multivariable Cox model built upon the IPTW-weighted cohort, with results visualized via a forest plot. Except for the Bonferroni-corrected pairwise comparisons, a two-sided P < 0.05 was considered statistically significant.

3 Results3.1 Patient characteristics and balance after weighting

A total of 142 patients with oligometastatic esophageal squamous cell carcinoma were included in this study, comprising 27 (19.0%) in the RT-first group, 87 (61.3%) in the Concurrent group, and 28 (19.7%) in the Sequential group. The baseline demographic and clinical characteristics for all three groups are detailed in Table 1. The study cohort exhibited substantial data maturity at the time of analysis. Across the entire population (N = 142), a total of 94 deaths (66.2%) and 86 progression events (60.6%) were documented during the follow-up period. Specifically, the number of death events was 21 (77.8%) in the RT-first group, 52 (59.8%) in the Concurrent group, and 21 (75.0%) in the Sequential group. For progression-free survival, the number of events was 18 (66.7%), 51 (58.6%), and 17 (60.7%), respectively.

CovariateRT-firstConcurrentSequentialSMDn278728Sex = Female (%)5 (18.5)24 (27.6)9 (32.1)0.211Age_Group = >=60 (%)22 (81.5)68 (78.2)16 (57.1)0.364ECOG = 1-2 (%)24 (88.9)81 (93.1)21 (75.0)0.342Tumor_Location (%)0.167Upper5 (18.5)15 (17.2)6 (21.4)Middle14 (51.9)51 (58.6)13 (46.4)Lower8 (29.6)21 (24.1)9 (32.1)Oligo_Type = Metachronous (%)14 (51.9)42 (48.3)18 (64.3)0.218Met_Organ_Num = 2-3 (%)4 (14.8)13 (14.9)6 (21.4)0.115Met_Lesion_Num = 4-5 (%)4 (14.8)10 (11.5)4 (14.3)0.066RT_Site = Oligometastatic (%)17 (63.0)40 (46.0)10 (35.7)0.374NLR (median [IQR])2.83 [2.40, 3.42]2.48 [1.96, 3.98]2.32 [1.62, 3.63]0.238

Baseline characteristic table of patients before weighting.

Prior to weighting, significant imbalances were present across the groups for several key covariates. Notably, the proportion of patients with an ECOG performance status of 1–2 was lower in the Sequential group (75.0%) compared to the Concurrent (93.1%) and RT-first groups (88.9%) (SMD = 0.342). Imbalances (all SMD > 0.2) were also observed for patient age, sex, type of oligometastasis, radiotherapy site, and baseline NLR. These initial imbalances highlight the potential for significant “confounding by indication” bias within this retrospective cohort.

To select the optimal propensity score model, we compared the covariate balancing performance of the multinomial logistic regression and GBM models. Table 2 summarizes the SMD values before and after weighting, while Figure 2 provides a visual comparison.

CovariateUnweighted SMDSMD Post-Weighted by GBMSMD Post-Weighted by LRECOG0.3420.2560.009RT_Site0.3740.4300.422Age_Group0.3640.3460.243NLR0.2380.1850.160Oligo_Type0.2180.0900.021Sex0.2110.2520.230Tumor_Location0.1670.1890.167Met_Organ_Num0.1150.1880.169Met_Lesion_Num0.0660.0800.098

Comparison of standardized mean differences for core covariates before and after weighting by two models.

Balance of covariates before and after weighting comparison between GBM and logistic regression models. GBM-Weighted used Gradient Boosting Machine model; LR-Weighted used Multinomial Logistic Regression model.

The results demonstrated that the logistic regression model achieved superior performance in balancing the key confounding factors. For instance, after weighting with the LR model, the SMD for ECOG performance status was reduced from 0.342 to a near-perfect 0.009, and the SMD for the type of oligometastasis decreased from 0.218 to 0.021. In contrast, the GBM model failed to adequately balance several covariates, including ECOG performance status (SMD = 0.256) and radiotherapy site (SMD = 0.430). Given that the traditional logistic regression achieved superior overall covariate balance and maintained model parsimony for our specific dataset, it was selected over the non-parametric GBM model to ensure the validity of the IPTW weights.

Therefore, all subsequent analyses were conducted using the stabilized IPTW calculated from the logistic regression model.

The follow-up for this study was conducted until August 2025, with a median follow-up time of 21.6 months (95% CI, 19.8–23.4 months). The disease control rate (DCR) was 66.7% in the RT-first group, 58.6% in the Concurrent group, and 60.7% in the Sequential group. There was no statistically significant difference in DCR among the three groups (χ² = 0.559, P = 0.756).

The final weighted multivariate Cox regression analysis revealed a significant difference in overall survival (OS) among the three groups (weighted log-rank test, P = 0.018). The weighted Kaplan-Meier curves demonstrated superior OS in the Concurrent group (Figure 3). The median OS was 22.7 months (95% CI, 18.2–35.6) in the Concurrent group, compared to 21.0 months (95% CI, 15.8–27.5) in the RT-first group and 18.1 months (95% CI, 12.0–24.2) in the Sequential group.

OS curves for three groups of patients.

There was no significant difference in PFS among the three groups (weighted log-rank test, P = 0.314) (Figure 4). The median PFS was 16.4 months (95% CI, 11.2–not reached) for the Concurrent group, 12.2 months (95% CI, 8.9–22.6) for the RT-first group, and 15.4 months (95% CI, 10.0–17.8) for the Sequential group.

PFS curves for three groups of patients.

A significant overall difference in OS was observed among the treatment groups (global Wald test, P = 0.02). Pairwise comparisons revealed that the Concurrent group had a significantly lower risk of death compared to the Sequential group (aHR = 0.47; 95% CI, 0.27–0.83; P’ = 0.010), a difference that remained significant after Bonferroni correction (P’ < 0.0167). Although a strong trend toward improved survival was also observed for the Concurrent group compared to the RT-first group (aHR = 0.58; 95% CI, 0.37–0.92; P’ = 0.022), this difference did not meet the significance threshold after Bonferroni correction (Table 3).

CharacteristicaHR95% CIP-valueOSTreatment GroupConcurrent vs. RT-first (reference)0.580.37 - 0.920.022Sequential vs. RT-first (reference)1.220.65 - 2.290.526Sequential vs. Concurrent (reference)2.111.20 - 3.700.010RT SiteOligometastasis vs. Primary lesion (reference)1.070.71 - 1.630.734PFSTreatment GroupConcurrent vs. RT-first (reference)0.700.40 - 1.23 0.216Sequential vs. RT-first (reference)0.960.47 - 1.94 0.899RT SiteOligometastasis vs. Primary lesion (reference)0.970.63 - 1.510.903

Weighted multivariable Cox regression analysis of the impact of treatment sequencing on OS and PFS.

To identify the independent prognostic factors for overall survival, we first conducted a univariate weighted Cox regression analysis on 10 potential prognostic variables. Based on a pre-specified screening criterion of P < 0.20, a total of six variables were selected as candidates for the multivariate model: treatment group, sex, age group, ECOG performance status, tumor location, and baseline NLR (Table 4).

CharacteristicUnivariable analysisMultivariable analysisPHR (95%CI)PaHR (95%CI)Treatment Group0.009<0.001RT-firstReferenceReferenceConcurrent0.0210.58 (0.36 - 0.92)0.0070.53 (0.33 - 0.84)Sequential0.5461.21 (0.65 - 2.28)0.4471.29 (0.67 - 2.50)Age0.0270.59 (0.37 - 0.94)0.0280.59 (0.37 - 0.94)NLR0.0611.01 (1.00 - 1.03)0.0021.02 (1.01 - 1.03)Tumor Location0.3000.030UpperReferenceReferenceMiddle0.1381.51 (0.88 - 2.61)0.0371.93 (1.04 - 3.58)Lower0.1951.54 (0.80 - 2.98)0.0741.96 (0.94 - 4.11)Sex0.0881.46 (0.95 - 2.24)––ECOG0.0490.57 (0.33 - 0.99)––

Univariable and multivariable weighted Cox regression analysis for OS.

Variables for the multivariate model were selected via backward elimination from candidate variables with a p-value < 0.20 in the univariate analysis. The final multivariate model included treatment group, age group, NLR, and tumor location. A dash (—) indicates that the variable was not included in the final multivariate model.

Subsequently, these six candidate variables were included in a multivariate weighted Cox model and subjected to a backward stepwise selection process with a removal criterion of P > 0.10. During this process, ECOG performance status (P = 0.401 in the full model) and sex (P = 0.252 in a subsequent model) were sequentially eliminated from the model.

In the final weighted multivariate Cox regression analysis for prognosis, four variables were identified as independent predictors of overall survival (Figure 5).

Forest plot of the final multivariable Cox regression model for independent prognostic factors of OS.

Concurrent treatment timing remained a strong protective factor (vs. the RT-first group; aHR = 0.53, 95% CI, 0.33–0.84; P = 0.007). Additionally, a higher baseline NLR (aHR per unit increase = 1.02; 95% CI, 1.01–1.03; P = 0.002) and a tumor location in the middle thoracic esophagus (vs. upper thoracic; aHR = 1.93; 95% CI, 1.04–3.58; P = 0.037) were independent risk factors for mortality. Interestingly, an age of ≥60 years was also found to be an independent protective factor (aHR = 0.59; 95% CI, 0.37–0.94; P = 0.028).

3.2 Safety analysis

The treatment regimens were well-tolerated by all patients in the study cohort. No patients discontinued therapy due to intolerable adverse events. All observed treatment-related adverse events (TRAEs) were mild to moderate in severity (predominantly Grade 1–2) and were manageable with timely intervention.

The most common toxicities included myelosuppression, pneumonitis, and abnormal liver or renal function. There was no statistically