Esophageal cancer (EC) is a major malignant tumor of the digestive system, ranking as the eighth most common cancer and the sixth leading cause of cancer-related mortality worldwide, with an estimated 604,000 new cases and 544,000 deaths annually (1, 2). Under the influence of genetic factors and various environmental factors, the incidence and mortality rates of EC exhibit variations across different geographical regions, genders, and ethnicities (3). Particularly remarkable is China, where esophageal squamous cell carcinoma (ESCC) constitutes the most prevalent subtype (approximately 90% of all esophageal cancer cases), and accounts for more than half of the new cases and deaths worldwide (4, 5).

Early-stage EC is primarily treated with endoscopic therapy and surgical resection (6). However, many patients are unsuitable for surgery because of the location in the upper thoracic segment of the esophagus, and many others are initially diagnosed at an advanced stage and thus lose the opportunity for surgery (6). Thereby, the standard systemic treatment, mainly encompassing chemotherapy and immunotherapy, is particularly crucial for unresectable or advanced EC. Prior to 2017, paclitaxel, platinum and fluorouracil were commonly utilized chemotherapeutic agents that dominated in the treatment for a long time, but the five-year survival rate remained below 5%, and even lower for advanced ESCC patients (7, 8). The treatment landscape was fundamentally transformed by the introduction of immune checkpoint inhibitors (ICIs), which target the programmed death 1 (PD-1)/programmed death‐ligand 1 (PD-L1) axis to enhance T-cell-mediated anti-tumor immunity (9). Since 2017, there have been continuous developments in the application of immunotherapy for advanced ESCC. The clinical exploration of immunotherapy began with the ATTRACTION-1 trial of nivolumab and the KEYNOTE-181 trial of pembrolizumab, which focused on its application in second-line and subsequent-line treatments (10–12). Then this success was subsequently extended to the first-line setting, with large-scale Phase III clinical trials (such as KEYNOTE-590 trial of pembrolizumab, CheckMate-648 trial of nivolumab, ORIENT-15 trial of sintilimab, and RATIONALE-306 trial of tislelizumab) demonstrating the significant efficacy of ICIs (13–16). Based on the results of these studies, both the 2023.V3 National Comprehensive Cancer Network (NCCN) Guidelines and the Chinese American Society of Clinical Oncology (ASCO) Guidelines for 2023 recommend the use of various immunotherapy regimens in combination with chemotherapy as first-line or subsequent-line treatment options for ESCC (7, 17, 18). However, the initiation of ICIs can result in the emergence of immune-related adverse events (irAEs), which represent a double-edged sword in cancer immunotherapy. On one hand, irAEs can affect multiple systems and are potentially life-threatening (19–23). On the other hand, their underlying pathophysiological mechanisms, including T-cell-mediated cytotoxicity, elevated autoantibody levels, inflammatory cytokine release, and complement-mediated inflammation, may reflect a robust anti-tumor immune response, thereby contributing to favorable therapeutic outcomes (20, 24). Accumulating evidence indicates that irAEs serve as a biomarker for improved outcomes in various malignancies, including lung cancer, hepatocellular carcinoma, gastric cancer, and renal cell carcinoma (19, 20, 23, 25–34). Research is now moving beyond the binary presence of irAEs to investigate the prognostic impact of single-organ irAE (uni-irAE), multiple-organ irAEs (multi-irAEs), and organ-specific irAE (26–30, 35). While the occurrence of irAEs predicts better outcomes in EC patients, the relationship between subtypes of irAEs (uni-irAE, multi-irAEs, and organ-specific irAE) and outcomes remains particularly underexplored (36–38).

Therefore, we conducted this study to determine how irAE characteristics, such as number, affected organ, and severity, influence therapeutic efficacy and prognosis in ESCC patients undergoing ICI therapy. Furthermore, we investigated the specific irAE co-occurrence patterns across multiple organs.

The occurrence of irAEs appears to be a promising predictive biomarker for ICI efficacy across various cancer type (19, 20, 23, 25–34). Previous studies have demonstrated that in patients with advanced EC, the development of irAEs was associated with significantly improved outcomes (23, 36). We performed a deeper analysis that found a more nuanced relationship, which reveals that uni-irAE served as an independent protective factor for both PFS and OS. Multi-irAEs were also independently associated with improved OS, though their benefit for PFS did not reach statistical significance possibly due to limited follow-up or sample size. Additionally, no significant difference in survival was observed between the uni-irAE and multi-irAEs groups. Overall, the occurrence of both uni-irAE and multi-irAEs predicted superior survival, consistent with findings in other malignancies (23, 27, 40–42).

The positive correlation between irAEs and treatment efficacy is thought to stem from shared immunological pathways. One such pathway is T cell cross-reactivity against shared antigens, supported by the discovery of identical T cell receptor (TCR) sequences in both tumors and irAE-affected tissues (43–49). Concurrently, cytokines play a significant role in both inducing irAEs and eliminating tumor cells. Patients with irAEs exhibit elevated levels of cytokines, such as Interferon-gamma (IFN-γ) and Interleukin-7 (IL-7), which are associated with both autoimmune toxicity and a favorable therapeutic response (50–55). Furthermore, ICI-induced depletion of regulatory T cells (Tregs) can unleash autoreactive T cells, simultaneously causing irAEs and enhancing anti-tumor responses (48, 56, 57). Finally, the composition of the gut microbiota is also known to shape both therapeutic efficacy and the likelihood of irAEs. Mechanistically, specific bacterial signatures can modulate systemic immunity and lower the T-cell activation threshold, thereby enhancing anti-tumor responses while simultaneously increasing susceptibility to autoimmune toxicity (44, 58–63). The prognostic value of multi-irAEs varies significantly across cancer types, being superior in gastric, small cell lung cancer, and renal cell carcinoma, but not in non-small cell lung cancer (23, 25–28). In our ESCC cohort, multi-irAEs were associated with the longest OS but an intermediate PFS, which was shorter than that of the uni-irAE group but longer than that of the non-irAE group. The development of multi-irAEs suggests a more robust systemic immune response or a profoundly altered tumor microenvironment that better facilitates immunotherapy (64).This heightened immune reactivity is believed to be the biological basis for the “tail effect,” which enables long-term survival even after initial progression (65). Conversely, the intermediate PFS may be attributable to the heterogeneity of tumor types or the limited sample size. Therefore, further prospective trials and immunological studies are warranted to delineate these complex relationships.

Our deeper analysis of irAEs revealed that co-occurrence patterns and severity are key prognostic factors, shifting the focus from a binary view to a more nuanced understanding.

The most frequent co-occurrences were thyroid-adrenal and thyroid-myocarditis pairs, and the time to onset for any irAE was similar between uni-irAE and multi-irAEs groups, indicating that the development of multiple irAEs is not time-dependent. While endocrine irAEs predicted the development of multi-irAEs, they were also associated with superior survival outcomes. Regarding the clinical timing of these prognostic markers, in some cases, such as skin irAE, it can serve as a clinical response marker earlier than the first imaging evaluation because its average time of onset (1.9 months) precedes the first imaging assessment. In other cases, such as hepatic (3.68 months), endocrine (4.17 months), and pulmonary toxicity (4.92 months), their average time of occurrence is indeed later than the first imaging evaluation. However, traditional imaging evaluation can no longer fully and accurately reflect the true efficacy of immunotherapy, particularly in tumors where cancer cell reduction and corresponding lymphocyte infiltration occur without changes in tumor volume. Therefore, even irAEs that occur later than imaging evaluation still possess auxiliary predictive functions.

The underlying mechanisms for irAE co-occurrence and the positive prognostic impact of endocrine irAEs remain unclear, but several mechanisms have been proposed. One proposed mechanism for multi-organ co-occurrence is a systemic, uncontrolled immune response, such as a cytokine storm, which can cause simultaneous damage across different systems (66–68). B cell activation may be another mechanism, as it elevates autoantibodies that can deposit in specific organs and trigger inflammation. For example, patients with hepatic irAEs showed increased levels of anti-thyroglobulin Immunoglobulin G (IgG), a B cell-derived antibody that can precipitate thyroid irAEs (19). The survival advantage of endocrine irAEs may stems from two key factors. First, given that endocrine tissues are known to exhibit intense lymphocyte and cytokine infiltration, they become a primary target in this systemic assault, leading to the onset of endocrine irAEs (69, 70). Second, their typically low-grade nature allows for management with gentle therapies, avoiding high-dose corticosteroids that would suppress the beneficial immune activity. A small sample size may explain the inconclusive findings for hepatic, pulmonary, and gastrointestinal irAEs.

Analysis of irAE severity revealed that mild irAEs were associated with superior DCR, PFS, and OS compared to non-irAE in ESCC patients, which corroborated early findings in lung cancer and renal cell carcinoma (42, 71–73). The prognosis of severe irAEs, however, is controversial; while some studies link them to the worst outcomes, others show no difference from non-irAE (74). In our cohort, severe irAEs showed a trend toward reduced OS when compared to mild irAEs, but they displayed the extension of OS than that of non-irAE. The inferior prognosis of severe irAEs compared to mild irAEs may be attributed to two primary factors. First, ICI treatment is often discontinued following the onset of severe irAEs, potentially limiting the cumulative benefits of continued therapy (75). Second, the high-dose corticosteroids required to manage severe irAEs may compromise the efficacy of immunotherapy (76–83). Then, the favorable prognosis of patients with severe irAEs comparing to the non-irAE group may be explained by the following mechanisms. First, the onset of any irAE can serve as a biomarker for a robust therapeutic response, as discussed previously. This enhanced immunity is thought to drive the “tail effect,” a phenomenon where the immune system continues to exert an anti-tumor effect even after treatment cessation (65, 84). Second, close clinical monitoring and prompt intervention for irAEs can prevent severe events from becoming life-threatening, thereby improving overall outcomes. Future studies with larger sample sizes are warranted to validate these findings.

The impact of glucocorticoids (GCs) on the efficacy of immunotherapy remains a subject of debate, potentially influenced by complex interplays between dosage, treatment duration, and administration timing. In this study, after excluding non-immunosuppressive doses (such as antiemetic dexamethasone or replacement hydrocortisone), the duration of GC treatment ranged from 7 to 73 days in 17 irAE patients. ROC analysis (AUC = 0.65) identified the GC treatment time longer than 54.5-day associated with shorter Median OS (8.65 vs. 28.17 months, p = 0.006) in these ESCC patients. The Median PFS also reflected a trend of harm associated with prolonged steroid use, although statistically non-significant (7.98 vs. 11.08 months, p = 0.115) (data not shown).

Currently, research on the prognostic impact of GCs primarily focuses on administration timing and dosage. Regarding timing, baseline use of high-dose GCs is often associated with poorer outcomes, likely reflecting baseline characteristics such as high tumor burden or poor performance status (80, 85, 86). In contrast, therapeutic intervention during the induction phase, typically the first 4 to 8 weeks, remains a subject of debate. Some studies suggest that early immunosuppression may impair long-term efficacy by interfering with T-cell clonal expansion and priming, potentially abrogating the initial anti-tumor immune response (81, 87). Conversely, other evidence indicates that patients requiring corticosteroids for early-onset irAEs still exhibit superior survival outcomes compared to those steroid naïve group (88).Subsequent large-scale inquiries and meta-analyses are still needed. Furthermore, the duration of the tapering period is critical; rapid tapering of less than 4–6 weeks can easily trigger irAE rebound, leading to cumulative harm from secondary high-dose exposure (83, 89). In our analysis, 54.5 days (approximately 8 weeks) emerged as a critical threshold for treatment duration. Finally, concerning dosage, evidence suggests that the peak dose is more immunologically disruptive than the cumulative dose, as ultra-high concentrations can induce lymphocyte apoptosis, leading to poorer outcomes (79, 90, 91). Therefore, in clinical practice, clinicians should aim to balance effective toxicity control with antitumor efficacy. We propose managing the intervention timing and tapering schedule to keep the total duration ideally within this 54.5-day window, while also carefully considering peak dosage to minimize immunological disruption.

This study has several limitations. First, its retrospective, single-center design and relatively small sample size limit our ability to account for confounding variables and increase the risk of selection bias. Second, variations in patient treatment regimens could have influenced the final outcomes. Third, the short follow-up period for some patients and the presence of censored data may have led to an incomplete assessment of median survival times. Considering these limitations, future follow-up studies, multi-center collaborations and prospective trials are warranted to enhance the reliability and validity of these findings.

The occurrence of both uni-irAE and multi-irAEs was associated with favorable prognosis in ESCC patients treated with ICIs. Furthermore, patients who developed endocrine irAEs or mild irAEs also demonstrated improved efficacy, suggesting their potential as clinical response markers for a positive response to therapy. This finding emphasizes the necessity of vigilant monitoring and early intervention for irAEs in patients undergoing ICIs.