Genomic co-alterations and immune-related adverse events shape outcomes of thymic malignancies treated with immune checkpoint inhibitors

Abstract

Thymic carcinoma, thymoma, and thymic neuroendocrine tumors are rare malignancies with limited treatment options in advanced stages. Immune checkpoint inhibitors (ICIs) have shown preliminary activity in thymic tumors, although their clinical benefit and toxicity profile remain incompletely defined. We retrospectively analyzed patients with thymic tumors treated with ICIs at MD Anderson Cancer Center from 2010 to 2024. Clinical characteristics, molecular profiles, treatment details, immune-related adverse events (irAEs), and survival outcomes were collected. The primary objective was ICI safety. Secondary endpoints included objective response rate (ORR), disease control rate (DCR), progression-free survival (PFS), and overall survival (OS). Cox regression identified predictors of survival. Forty-two patients were analyzed (median age, 59 years; 55% male), comprising 29 thymic carcinomas, 8 thymomas, and 5 neuroendocrine tumors. ICIs were administered as monotherapy (55%) or combinations (45%), most commonly anti-PD-1 agents (86%). IrAEs occurred in 60% overall and in all patients with thymoma. Grade ≥3 irAEs developed in 19%, including one treatment-related death from pneumonitis. Common irAEs included fatigue, rash, and musculoskeletal symptoms. Notable neuromuscular syndromes included concurrent myasthenia gravis and myocarditis in thymoma. Among 37 RECIST-evaluable patients, ORR was 21%, and DCR was 78%. Median PFS was 6.5 months, and median OS was 9.1 months. Any-grade irAEs were associated with reduced mortality risk (HR, 0.30). TP53 mutations and lung metastases predicted worse OS, whereas CDKN2A alterations correlated with improved OS. ICIs demonstrated preliminary activity in thymic tumors but were associated with substantial toxicity, especially in thymoma. Biomarker-driven patient selection and vigilant irAE monitoring are needed.

Introduction

Thymic tumors are biologically heterogeneous and clinically challenging malignancies arising from the thymus, an organ central to T-cell maturation and immune tolerance¹. Thymic tumors comprise 3 distinct subtypes: thymoma, thymic carcinoma, and thymic neuroendocrine tumors. Collectively, they account for less than 1% of all adult cancers¹. Treatment of advanced disease has substantial therapeutic challenges due to the tumors’ immune-rich microenvironment and frequent association with autoimmune diseases and paraneoplastic syndromes².

Thymoma is the most common subtype of thymic epithelial tumor, characterized by lobulated architecture with fibrous septa and a mixture of neoplastic epithelial cells and immature T lymphocytes. Thymomas are strongly associated with autoimmune diseases, particularly myasthenia gravis, pure red cell aplasia, and hypogammaglobulinemia¹.

Thymic carcinoma is an aggressive epithelial malignancy that lacks the organotypic features of thymoma. Histologically, it displays pronounced cytologic atypia and infiltrative growth and often resembles squamous cell carcinoma¹. Thymic carcinomas have a higher incidence of metastasis at presentation and a worse prognosis than thymomas. Furthermore, unlike thymomas, thymic carcinomas are rarely associated with autoimmune diseases, suggesting a fundamentally distinct tumor-immune microenvironment.

Neuroendocrine tumors of the thymus, including atypical carcinoids and large-cell neuroendocrine carcinomas, represent the least common subtype of thymic malignancy³. These tumors often affect middle-aged men, are invasive, and may be associated with endocrinopathies such as Cushing syndrome and acromegaly³. Histopathology shows neuroendocrine differentiation and high mitotic activity.

The Masaoka-Koga system has historically been the most widely used staging framework for thymic epithelial tumors (Table S1), classifying disease by local invasion and metastatic spread from encapsulated stage I tumors to stage IV disease with pleural/pericardial dissemination (IVA) or lymphogenous/hematogenous metastasis (IVB)⁴. Although prognostically informative, its reliance on capsular invasion and its development primarily in thymoma and thymic carcinoma limits its applicability to thymic neuroendocrine tumors. More recently, the International Thymic Malignancies Interest Group and the International Association for the Study of Lung Cancer proposed a TNM-based classification system that was incorporated into the AJCC 8th edition staging system for thymic malignancies⁵. This framework provides a standardized approach across thymic tumor subtypes by classifying disease according to primary tumor extent, nodal involvement, and metastatic spread. In the present study, Masaoka-Koga staging was retained for historical comparability, while AJCC 8th edition TNM stage grouping was also abstracted when available to better characterize disease extent using the contemporary unified staging framework

The immune microenvironment of thymic tumors is uniquely complex. Thymomas have functional thymic epithelium capable of shaping the T-cell repertoire, facilitating the escape of autoreactive clones and predisposing to autoimmunity⁶. These tumors typically exhibit dense lymphocytic infiltration and high programmed death-ligand 1 (PD-L1) expression, features that suggest immunogenicity and susceptibility to immune-mediated toxic effects⁷. Thymic carcinomas, though generally devoid of lymphoid stroma, can demonstrate PD-L1 upregulation and tumor-infiltrating lymphocytes, indicating potential sensitivity to immune checkpoint inhibitors (ICIs)⁶. However, both thymomas and thymic carcinomas are typically characterized by low tumor mutational burden, limiting neoantigen diversity⁸. Neuroendocrine tumors of the thymus, once considered immunologically quiescent, may display an immune-inflamed phenotype with PD-L1 expression on tumor cells⁷ and CD4⁺ and CD8⁺ T-cell infiltration, although programmed cell death protein 1 (PD-1) expression on lymphocytes is rare, suggesting impaired T-cell activation⁹.

Over the past decade, ICIs targeting PD-1, PD-L1, and cytotoxic T lymphocyte–associated antigen-4 have transformed cancer therapy¹⁰. However, their clinical activity in thymic tumors has been modest, with durable benefit confined to a subset of patients^11–14 with increased susceptibility to severe immune-related adverse events (irAEs), particularly myocarditis, myositis, and hepatitis^12–14. Although clinical trials have provided important insights into the efficacy and toxicity of ICIs in thymic malignancies, the true incidence, spectrum, and clinical implications of irAEs in real-world settings remain incompletely characterized.

Consequently, the overall risk-benefit profile of ICIs in thymic malignancies remains poorly defined. Furthermore, the lack of integrated analyses incorporating histologic subtype, molecular features and immunologic correlates has limited the identification of biomarkers that may guide patient selection and optimize therapeutic outcomes. Real-world studies evaluating the interplay between tumor biology, treatment efficacy and irAEs are therefore needed.

We hypothesized that distinct clinical and molecular features may influence treatment outcomes in patients with thymic tumors receiving ICIs. To test this hypothesis, we conducted a comprehensive retrospective analysis of patients with thymic malignancies treated with ICIs, evaluating clinical characteristics, molecular biomarkers, treatment outcomes and irAEs to identify factors associated with overall survival.

Results

Patient eligibility and study population

We screened a total of 74 patients with thymic tumors who received at least 1 dose of an ICI at MD Anderson from 2010 to 2024 for study eligibility. We excluded 32 patients: 28 had received ICIs exclusively as part of an ongoing clinical trial (NCT03295227), and 4 had incomplete treatment records. The final cohort consisted of 42 patients (Fig. S1).

Patient characteristics

Among the 42 study patients (Table 1), the median age at initiation of ICI therapy was 59 years (range, 22–79 years), and 55% were male. The racial distribution was predominantly White (62%), followed by Black/African American (21%), Asian (10%), and other (7%). Most patients had an Eastern Cooperative Oncology Group performance status of 0–1 (91%). Three patients had pre-existing autoimmune disease: one had myasthenia gravis, another had rheumatoid arthritis, systemic lupus erythematosus, and Sjögren syndrome, and the third had ulcerative colitis.

Table 1.

Baseline clinical characteristics of the patient cohort (n = 42)

Characteristic	n (%)
Median age, years (range)	59 (22–79)
Sex
Male	23 (55)
Female	19 (45)
Race
White	26 (62)
Black/African American	9 (21)
Asian	4 (10)
Other	3 (7)
ECOG performance status
0	7 (17)
1	31 (74)
2	3 (7)
3	1 (2)
Pre-existing autoimmune history
Yes	3 (7)
Myasthenia gravis	1 (2)
Ulcerative colitis	1 (2)
Rheumatoid arthritis/SLE/Sjögren syndrome	1 (2)
Tumor subtype
Thymic carcinoma	29 (69)
Thymoma	8 (19)
Neuroendocrine tumor of the thymus	5 (12)
TNM stage
IVa	12 (29)
IVb	30 (71)
Sites of metastasis
Lymph nodes	31 (74)
Pleura	25 (60)
Bone	17 (40)
Liver	13 (31)
Lungs	11 (26)
Prior lines of systemic therapy
Median (range)	1 (0–5)
Chemotherapy	37 (88)
Targeted therapy	11 (26)
Other therapeutic modalities
Radiotherapy	31 (74)
Surgery	19 (45)

Baseline demographic, clinical, and treatment characteristics of the 42 patients with thymic tumors treated with immune checkpoint inhibitors. ECOG Eastern Cooperative Oncology Group, SLE systemic lupus erythematosus.

Histologically, thymic carcinoma was the most common tumor subtype (n = 29 [69%]), followed by thymoma (n = 8 [19%]) and thymic neuroendocrine tumors (n = 5 [12%]). Most patients had Masaoka stage 4b disease (n = 30 [71%]). AJCC 8th edition TNM staging classified 12 patients as stage IVA and 31 patients as stage IVB. The most common metastatic sites were the lymph nodes (74%), pleura (59%), bone (41%), liver (31%), and lungs (33%). The median number of metastatic sites was 3 (range, 1–5).

Patients received a median of 1 prior line of systemic therapy (range, 0–5). Most patients had received chemotherapy (88%) and radiotherapy (74%), and 45% underwent prior surgical resection.

Physicians administered ICIs as monotherapy in 55% of the patients and in combination regimens in 45% of them, including with chemotherapy (21%), other ICIs (14%), targeted therapy (5%), radiotherapy (2%), and an investigational oncolytic virus (2%). The majority of patients received anti-PD-1 agents (86%), comprising pembrolizumab (n = 24) and nivolumab (n = 10). Additional agents included anti-cytotoxic T lymphocyte–associated antigen-4 (5%), anti-LAG-3 (5%), and anti-PD-L1 (2%) antibodies and an anti-PD-1/IL-2 fusion antibody (2%).

Molecular analysis

PD-L1 expression was assessed via immunohistochemistry using the CPS in 23 patients, 38% of whom had a CPS of at least 1 (Table 2). Among the 10 patients tested through next-generation sequencing, none had a tumor mutational burden greater than 10 mutations per megabase or microsatellite instability–high status. Next-generation sequencing results were available for 34 patients, with the most common genomic alterations observed being those of TP53 (17%), CDKN2A (17%), and KIT (17%). CDKN2A alterations were observed exclusively in thymic carcinoma and were associated with PD-L1 expression ≥1 in 57% of evaluable cases.

Table 2.

Biomarkers and genomic alterations in the patient cohort (n = 42)

Molecular testing	n (%)
PD-L1 CPS ≥ 1 (tested in 23 patients)	11 (48)
Genomic alterations
CDKN2A	7 (17)
KIT	7 (17)
TP53	7 (17)
BAP1	3 (7)
BCOR	3 (7)
PIK3CA	3 (7)
SMARCA4	3 (7)
TET2	3 (7)
ATM	2 (5)
CREBBP	2 (5)
CYLD	2 (5)
KRAS	2 (5)
MAPK	2 (5)
MET	2 (5)
RET	2 (5)

Programmed death-ligand 1 (PD-L1) expression and genomic alterations identified in the study cohort of patients with thymic tumors treated with immune checkpoint inhibitors. Genomic alterations were identified using Clinical Laboratory Improvement Amendments (CLIA)-certified next-generation sequencing assays.

Safety

IrAEs occurred in 60% of the patients overall. All patients with thymoma experienced irAEs, compared with 52% of those with thymic carcinoma and 40% of those with thymic neuroendocrine tumors (Table S2). Severe irAEs (grade ≥ 3) occurred in 19% of the patients (n = 8), affecting 50% of those with thymoma in particular. Severe irAEs did not predominantly occur in patients with documented pre-existing autoimmune disease. Only 1 of the 8 patients with severe irAEs had a history of autoimmune disease, specifically pre-existing myasthenia gravis. One treatment-related death occurred due to pneumonitis. Treatment was delayed or withheld in 19% of patients and permanently discontinued in 17%, and no dose reductions were reported. The most frequent irAEs were fatigue (19%), rash (19%), musculoskeletal symptoms (17%), hepatitis (10%), nausea (7%), myocarditis (7%), and myasthenia gravis–like symptoms (7%) (Table 3). Notably, 3 patients with thymoma had development of myasthenia gravis–like symptoms (2 de novo and 1 with pre-existing disease). Two of the patients with myasthenia gravis–like symptoms also experienced concurrent myocarditis. Both patients who developed myocarditis received corticosteroids, with subsequent resolution of symptoms. The median time from treatment initiation to onset of severe irAEs was 1.4 months (range, 0.3–14.4 months). Eight of nine severe irAEs occurred within the first 4 months of treatment. Most resolved with the use of corticosteroids (78%), although 2 patients experienced steroid-refractory events.

Table 3.

IRAEs in the study cohort (n = 42)

irAE	Grade 1–2	Grade ≥G3	n (%)
Fatigue	8	0	8 (19)
Rash	7	1	8 (19)
Musculoskeletal	5	2	7 (17)
Hepatitis	2	2	4 (10)
Myocarditis	1	2	3 (7)
Myasthenia gravis	1	2	3 (7)
Nausea	3	0	3 (7)
Thyroiditis	2	0	2 (5)
Pruritus	2	0	2 (5)
Flu-like symptoms	2	0	2 (5)
Pneumonitis	0	1	1 (2)
Colitis	0	1	1 (2)
Meningitis	0	1	1 (2)
Acute kidney injury	0	1	1 (2)
Neuropathy	0	1	1 (2)

Immune-related adverse events observed in the study cohort of patients with thymic tumors treated with immune checkpoint inhibitors. Adverse events were graded according to the Common Terminology Criteria for Adverse Events (CTCAE), version 5.0. Data are presented according to maximum reported grade. Patients may have experienced more than one irAE. irAE immune-related adverse event.

Efficacy

The median number of treatment cycles was 4, with a median treatment duration of 2.8 months (range, 0–27.0 months). Among 37 RECIST evaluable patients, the ORR was 21%, comprising 9 partial responses (8 in thymic carcinoma cases and 1 in a thymoma case) (Table 4). Notably, 5 of the 9 responders received anti-PD-1 therapy in combination with carboplatin and paclitaxel. Nearly half of the cohort (49%) experienced clinical benefit, primarily patients with thymic carcinoma (n = 12), with additional benefit in 5 thymoma cases and 1 thymic neuroendocrine tumor case. The DCR was 78%, including 20 patients who had stable disease. When stratified by regimen, the ORR was 18% with monotherapy and 33% with combination therapy, corresponding to DCRs of 68% and 93%, respectively (Table S3). Most patients (84%) who had any-grade irAEs experienced disease control (stable disease or partial response) with immune checkpoint inhibition. The median PFS duration was 6.5 months (95% CI, 4.0 months–not reached) (Fig. 1), and the median OS duration was 9.1 months (95% CI, 4.5–31.3 months) (Fig. 2).

Table 4.

Best objective response per tumor type

Response	Thymic carcinoma (n = 25)	Thymoma (n = 7)	Thymic neuroendocrine tumor (n = 5)	Total evaluable patients (n = 37)
PR	8	1	0	9 (24%)
SD	10	6	4	20 (54%)
PD	7	0	1	8 (22%)

Best objective response to immune checkpoint inhibitor therapy according to thymic tumor subtype among RECIST-evaluable patients. Tumor response was assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1. PD progressive disease, PR partial response, SD stable disease.

Fig. 1. PFS probability in the study cohort.

Kaplan–Meier curve of progression-free survival (PFS) in the overall study cohort of patients with thymic tumors treated with immune checkpoint inhibitors. Shaded areas represent 95% confidence intervals. Numbers at risk and cumulative events are shown below the x-axis.

Fig. 2. OS probability in the study cohort.

Kaplan–Meier curve of overall survival (OS) in the overall study cohort of patients with thymic tumors treated with immune checkpoint inhibitors. Shaded areas represent 95% confidence intervals. Numbers at risk and cumulative events are shown below the x-axis.

Durable response despite early treatment discontinuation for severe irAEs

One study patient in their 40s had metastatic thymoma involving the lungs and pleura achieved a durable response despite receiving only a single cycle of immune checkpoint inhibitor therapy. Prior to immunotherapy, the patient had received multiple lines of systemic treatment, including pemetrexed; cisplatin, and etoposide; and cyclophosphamide, doxorubicin, cisplatin, and prednisolone (Fig. S2). Nivolumab was initiated in October 2023 at a dose of 480 mg intravenously every 4 weeks but was discontinued after the first cycle because of grade 3 myocarditis and grade 3 myasthenia gravis. Despite the early treatment cessation, the patient achieved a partial response, with a 61% reduction in tumor burden according to RECIST version 1.1 crtieria. Notably, this response was durable and persisted for 14 months following treatment discontinuation.

Biomarker analyses

Clinical and molecular factors of OS were evaluated (Table 5). The development of any-grade irAEs correlated with improved outcomes, conferring a 70% reduction in the risk of death (HR, 0.30 [95% CI, 0.11–0.98]; P = 0.04). Landmark analyses at 3, 6, 9, and 12 months further reinforced this relationship. Patients who had an irAE before each time point experienced consistently longer post-landmark OS than did those without irAEs (Fig. S3). Although these differences did not reach statistical significance (log rank P = 0.19–0.56), likely owing to small sample size, hazard ratios for overall survival across all landmark analyses ranged from 0.26–0.45 and remained consistently below 1.0, indicating a numerically lower risk of death among patients who developed irAEs compared with those who did not. (Table S4). No significant association was observed between genomic alterations and the development of irAEs or severe irAEs.

Table 5.

Predictors of OS in the study cohort

OS predictor	HR (95% CI)	P-value
Any grade IRAEs	0.30 (0.11 - 0.98)	0.040
Lung metastasis	6.90 (2.40 -19.80)	<0.001
TP53 alterations	7.80 (2.20 - 27.80)	0.002
CDKN2A alterations	0.20 (0.04 - 0.54)	0.004

Clinical and molecular predictors of OS in the study cohort of patients with thymic tumors treated with immune checkpoint inhibitors. Hazard ratios, 95% confidence intervals, and corresponding P-values were derived from Cox proportional hazards regression analysis. CI confidence interval, HR hazard ratio, irAE immune-related adverse event, OS overall survival.

Among genomic biomarkers, CDKN2A alterations were associated with favorable OS, with an 80% reduction in mortality risk (HR, 0.2 [95% CI, 0.04–0.54]; P = 0.004). CDKN2A alterations were not associated with improved ORR. Adverse prognostic markers included lung metastasis (HR, 6.9 [95% CI, 2.4–19.8]; P < 0.001) and TP53 mutations (HR, 7.8 [95% CI, 2.2–27.8]; P = 0.002), both of which were linked with a markedly increased risk of death.

Discussion

Our study provided one of the most comprehensive characterizations to date of ICI outcomes in patients with thymic tumors, integrating clinical, pathologic, and molecular correlates of survival. In this rare disease context, in which therapeutic options remain limited and outcomes are generally poor, our findings highlight both the promise and challenges of immunotherapy.

IrAEs occurred frequently, affecting the majority of the patients, and were particularly pronounced in those with thymoma, all of whom developed irAEs. The toxicity spectrum was also distinctive: musculoskeletal manifestations, myasthenia gravis–like syndromes, and myocarditis emerged as hallmark events, with the last 2 frequently co-occurring. This highlights that patients with thymoma are uniquely susceptible to immune-mediated toxicity, underscoring the central role of the thymus in T-cell education and immune tolerance.

Mechanistically, neoplastic thymic epithelial cells often exhibit loss of autoimmune regulator gene expression, impairing negative selection and permitting autoreactive T cells to persist in the periphery¹⁵. This intrinsic defect is compounded in patients with thymoma, who commonly harbor pre-existing autoantibodies against acetylcholine receptor, titin, and ryanodine receptor, antigens that are expressed in both skeletal and cardiac muscle¹⁶. Shared antigenicity across neuromuscular and myocardial tissues facilitates antigen cross-recognition, whereas epitope spreading amplifies the autoimmune cascade by exposing secondary antigens during tissue injury¹⁷. Recent studies further highlighted how distorted thymic architecture, reduced major histocompatibility complex expression, and altered T-cell development compromise central tolerance, fostering autoreactive T- and B-cell responses that predispose patients with thymoma to immune-related toxicity¹⁸. Together, these mechanisms explain the striking clinical overlap of myasthenia gravis–like syndromes, myositis, and myocarditis, which frequently co-occur and can evolve in parallel. Consistent with this biology, prior trials with ICIs in patients with thymic tumors demonstrated high rates of severe irAEs, including hepatitis, myocarditis, and myasthenia gravis, with the highest incidence in thymoma cases^12–14. Similarly, our study demonstrated a substantial burden of irAEs among patients with thymoma. Although only two myocarditis events were observed, both occurred in patients with thymoma and concurrent myasthenia gravis–like symptoms. While this small number precludes definitive risk stratification, it supports heightened vigilance for cardiac irAEs when neuromuscular symptoms emerge in this subgroup. Because most severe irAEs occurred within the first 4 months of ICI treatment, early monitoring is particularly important. In patients with thymoma who develop neuromuscular or cardiopulmonary symptoms, prompt clinical assessment, electrocardiography, troponin testing, and early cardiology involvement should be considered.

Despite safety challenges, ICIs have produced meaningful clinical activity in a subset of patients with thymic tumors, with their efficacy varying by histologic subtype. In phase 2 monotherapy trials, pembrolizumab achieved an ORR of 28.6% (95% CI, 8.2–64.1%) and a DCR of 100% in thymoma cases, with a median PFS time of 6.1 months¹². In thymic carcinoma cases, pembrolizumab monotherapy attained ORRs ranging from 19% to 22%, DCRs ranging from 53% to 70%, and median PFS durations ranging from 4.2 to 6.1 months^12,13. Nivolumab monotherapy demonstrated comparable results, with an ORR of 12% (95% CI, 5–25%), DCR of 63% (95% CI, 48–77%), median PFS time of 6.0 months, and median OS time of 21.3 months in patients with type B3 thymoma or thymic carcinoma¹⁴. In our study, the ORR and DCR were 21% and 71%, respectively, across evaluable patients, aligning with prior phase 2 data. Notably, response outcomes favored combination therapy, which achieved an ORR of 33% and a DCR of 93%, compared with an ORR and DCR of 18% and 68%, respectively, for monotherapy. Nearly half of all responses occurred with anti-PD-1 therapy combined with chemotherapy, underscoring the potential synergy of combinatorial approaches. Supporting this, the phase 2 CAVEATT trial of avelumab plus axitinib demonstrated an ORR of 34% in patients with advanced thymic carcinoma or B3 thymoma, whereas the phase 2 MARBLE study of atezolizumab plus carboplatin and paclitaxel demonstrated an ORR of 56% in patients with thymic carcinoma^19,20. Collectively, these findings indicate that whereas anti-PD-1 monotherapy benefits a subset of patients, combination strategies may yield greater efficacy and warrant prioritization in future biomarker-driven studies.

Exploratory biomarker analyses identified both clinical and molecular correlates of survival. The association between any-grade irAEs and improved OS in our study aligns with evidence of increased therapeutic efficacy in cases of immune-related toxicity^21,22. In patients with thymic tumors, this link may be particularly pronounced given the central role of the thymus in T-cell selection and tolerance. The development of irAEs may therefore represent a surrogate biomarker of robust immune engagement, with increased T-cell activity directed against both tumor antigens and self-antigens. Although this association is encouraging, it must be interpreted with caution, as not all irAEs are beneficial, and high-grade irAEs can be life-threatening. Clinically, this finding underscores the importance of vigilant monitoring and early intervention to enable patients to derive therapeutic benefit from ICIs without experiencing irreversible harm.

The unexpected observation that CDKN2A alterations correlated with improved OS raises intriguing biologic hypotheses. CDKN2A mutations occurred exclusively in patients with thymic carcinoma, a subtype typically associated with aggressive clinical behavior and high mutational complexity. Truncating mutations, such as frameshift or nonsense mutations, result in the production of a shortened, nonfunctional protein. This impairs the function of the p16^INK4A and p14^ARF proteins, which are key regulators of the cell cycle, normally preventing uncontrolled cell division. Similarly, deletions of the CDKN2A gene, in which part or all of the gene is lost during DNA replication, lead to the loss of these tumor suppressor proteins, further promoting dysregulated cell proliferation. CDKN2A loss is classically linked with uncontrolled cell-cycle progression and poor survival outcomes²³.

CDKN2A-altered tumors in this study did not demonstrate higher response rates, but patients with them appeared to derive prolonged survival benefit from immune checkpoint blockade. The absence of an association between CDKN2A alterations and objective response rate suggests that the observed survival advantage may not be mediated solely through enhanced tumor shrinkage. Rather, CDKN2A loss may influence the durability of disease control or modulate the tumor-immune microenvironment in ways that prolong survival despite similar initial response rates. One potential explanation for this is that CDKN2A loss may induce replicative stress and genomic instability, thereby expanding the neoantigen repertoire and enhancing tumor immunogenicity^23,24. Alternatively, cell-cycle disruption may activate innate immune pathways such as cGAS-STING, amplifying tumor-immune interactions²⁵. Consistent with this, 57% of the patients harboring CDKN2A alterations in our cohort exhibited a PD-L1 CPS of at least 1, suggesting a potential association between CDKN2A alterations and features of an immune-responsive tumor phenotype that may increase susceptibility to immune checkpoint blockade.

Beyond its immunologic implications, CDKN2A loss frequently co-occurs with MTAP deletion, creating a potential therapeutic vulnerability targetable with PRMT5 inhibitors, which is a treatment approach warranting evaluation for thymic carcinoma. Moreover, cell-cycle pathway inhibition represents another rational strategy for thymic tumors with CDKN2A loss. In the phase 2 KCSG LU17-21 trial, palbociclib monotherapy produced an ORR of 12.5%, median PFS duration of 11.0 months, and median OS duration of 26.4 months in patients with pretreated thymic epithelial tumors, supporting the feasibility of cell cycle–directed therapy for this disease²⁶.

Notably, the influence of CDKN2A alterations on ICI response appears to be context-dependent, varying across tumor types. Whereas CDKN2A loss has been associated with adverse survival outcomes of urothelial carcinoma and has not demonstrated consistent predictive value for other malignancies, preclinical EGFR-mutant non-small cell lung cancer studies suggested that CDKN2A deficiency can sensitize tumors to PD-1 blockade via c-MYC–mediated PD-L2 upregulation and enhanced CD8⁺ T-cell activation^27,28. Whether these mechanisms extend to thymic carcinoma remains uncertain and warrants mechanistic validation in preclinical models and confirmation with larger, prospective datasets.

The presence of lung metastasis was strongly associated with poor survival in our study, reflecting both high disease burden and the unique biology of thoracic spread. The lung microenvironment is enriched with immunoregulatory macrophages, myeloid-derived suppressor cells, and regulatory T cells, all of which can create profound local immunosuppression²⁹. In addition, metastatic deposits within the lung may be less accessible to immune effector cells due to altered vascular permeability and stromal remodeling²⁹. These findings reinforce the clinical impression that patients with pulmonary involvement have a particularly aggressive disease trajectory and may need intensified or combinatorial treatment strategies. The findings also highlight the need for translational work interrogating the tumor-immune ecosystem of pulmonary metastases in patients with thymic tumors to inform rational therapeutic interventions.

TP53 mutation emerged as a potent adverse prognostic marker in our study, conferring a nearly 8-fold increase in risk of death. TP53 alterations are well-established drivers of genomic instability, aggressive biology, and treatment resistance across various tumor types and may contribute to immune evasion³⁰. Mechanistically, TP53 loss disrupts apoptotic signaling, impairs antigen presentation, and fosters a tumor microenvironment skewed toward immune suppression³⁰. The prognostic significance of TP53 alterations observed in our study aligns with prior molecular profiling efforts in thymic malignancies. TP53 is the most frequently altered genomic alteration in thymic carcinoma and metastatic thymoma, suggesting a role in tumor progression and dissemination³¹. Notably, it has been demonstrated that TP53-mutant thymic carcinomas exhibited significantly shorter DFS and OS³¹. Taken together, these findings support TP53 as a marker of aggressive tumor biology and raise the possibility that TP53-mutant thymic malignancies may represent a distinct molecular subgroup characterized by resistance to immune checkpoint blockade. Future therapeutic approaches for these patients may require biomarker-driven strategies incorporating p53 reactivation, DNA damage response targeting, or rational combination immunotherapy regimens³¹.

The retrospective study design inherently introduces bias, particularly in patient selection, treatment decisions, and documentation of outcomes, which may limit the generalizability of findings. Because we performed a retrospective, single-center analysis, our findings may not be fully generalizable to all practice settings. However, MD Anderson serves as a major referral center for patients with thymic tumors, and our cohort reflects real-world management of this rare, heterogeneous disease. In addition, the modest sample size in our study reduces the statistical power, increasing the risk of type II errors. Furthermore, the treatment heterogeneity in our study, with patients receiving different prior therapies and varied ICI regimens, complicates attribution of efficacy or toxicity to specific strategies. Molecular profiling was incomplete and performed in a subset of patients, raising the possibility of selection bias. Finally, PD-L1, tumor mutational burden, and microsatellite instability testing lacked uniformity across the cohort, which may have introduced variability in biomarker interpretation.

Despite these limitations, this study provides integrated clinical, safety, and molecular data for a disease for which prospective evidence of benefit with immune checkpoint inhibitors is exceedingly scarce. The high rate of irAEs observed in patients with thymoma highlights the unique immune susceptibility of this population. In contrast, signals of efficacy were observed in patients with thymic carcinoma. Additionally, exploratory analyses suggested that TP53 mutations may have prognostic significance in this tumor type. These findings, although preliminary, can help inform the design of prospective studies and serve as a benchmark for future multi-institutional collaborations in studying this rare malignancy.

ICIs demonstrated meaningful clinical activity in patients with thymic tumors, with the greatest benefit observed in those with thymic carcinoma. However, their use remains tempered by a distinctive and sometimes severe toxicity profile, particularly for thymoma. The occurrence of irAEs was associated with improved OS, whereas TP53 mutations and lung metastases were associated with markedly worse OS, underscoring the interplay between tumor biology and ICI efficacy. Importantly, CDKN2A alterations, found exclusively in thymic carcinoma cases, were linked with improved survival, suggesting a role for them as biomarkers of immune sensitivity and as therapeutically actionable targets. Given the frequent co-deletion of CDKN2A and MTAP, PRMT5 inhibition represents a rational strategy for future investigation in this subset of patients. Although exploratory, our findings generated testable hypotheses to guide biomarker development, refine patient selection, and inform the design of prospective, multi-institutional trials aimed at advancing precision immunotherapy for this rare and challenging disease.

Methods

Study design

This was a single-center, retrospective cohort study performed at The University of Texas MD Anderson Cancer Center. The index date was when the first dose of an ICI was administered, and patients were followed up until death or the last documented clinical contact. The data cutoff was December 31, 2024. All patient records were deidentified before analysis.

The protocol was reviewed by the Institutional Review Board of The University of Texas MD Anderson Cancer Center (IRB Registration No. IRB00000121) and was determined to be exempt from full IRB review under the category of secondary research involving existing data. The requirement for informed consent was waived because the study involved retrospective analysis of deidentified clinical data without direct patient contact. The study was conducted in accordance with the Declaration of Helsinki, the International Council for Harmonization Good Clinical Practice guidelines, and applicable regulatory requirements.

Population

Adults aged 18 years or older with a histologically confirmed thymic tumor (thymoma, thymic carcinoma, or thymic neuroendocrine tumor) were eligible if they had unresectable, recurrent, or metastatic disease not amenable to definitive therapy and had received at least 1 dose of an ICI (as monotherapy or in combination) from January 1, 2010, to December 31, 2024. Exclusion criteria were a concurrent malignancy requiring active treatment and clinical or imaging data insufficient for assessing outcomes. In addition, patients enrolled in an ongoing phase 1 study of pembrolizumab in the treatment of unresectable thymoma and thymic carcinoma (NCT03295227) were also excluded because trial-specific data-use restrictions precluded inclusion in the retrospective analysis.

Objectives

The primary objective of this study was to characterize the incidence and severity of irAEs in patients with thymic tumors receiving ICIs, which were graded using the Common Terminology Criteria for Adverse Events (CTCAE; version 5.0). Secondary objectives were evaluation of antitumor activity, defined by the objective response rate (ORR; proportion of patients experiencing a complete or partial response), clinical benefit rate (proportion of patients experiencing a complete or partial response or stable disease lasting ≥6 months), and disease control rate (DCR; proportion of patients experiencing a complete or partial response or stable disease) according to the Response Evaluation Criteria in Solid Tumors (RECIST; version 1.1). Survival outcomes included progression-free survival (PFS), defined as the time from initiation of ICI administration to radiographic progression or death, and overall survival (OS), defined as the time from initiation of ICI administration to death resulting from any cause. Patients who remained event-free or were alive at data cutoff were censored at their last documented follow-up visit. An exploratory aim was to identify clinical and molecular biomarkers associated with OS. PD-L1 expression was assessed via immunohistochemistry using the combined positive score (CPS). The CPS was defined as the number of PD-L1–stained tumor cells, lymphocytes, and macrophages divided by the total number of viable tumor cells, which was then multiplied by 100.

Data collection

Clinical and molecular variables were obtained from electronic health records and prospectively maintained institutional databases using a standardized case report form. Data were abstracted across 5 domains: demographics (age at ICI initiation, sex, race, Eastern Cooperative Oncology Group performance status), clinical characteristics (tumor subtype, Masaoka stage, metastatic sites, prior therapies), safety (irAE type and grade according to CTCAE (version 5.0]), efficacy (best overall response as per RECIST [version 1.1], PFS, OS), and biomarkers (PD-L1 expression according to standardized immunohistochemistry assays and genomic alterations identified in validated next-generation sequencing panels).

Statistical analysis

Descriptive statistics were used to summarize baseline characteristics and clinical variables. PFS and OS were estimated using the Kaplan-Meier method and reported as median values with 95% CIs. To identify predictors of OS, univariable and multivariable Cox proportional hazards models were constructed, incorporating clinical, molecular, and treatment-related covariates. Quantitative variables were analyzed as continuous measures, with covariates entered directly into Cox models. Hazard ratios (HRs) and 95% CIs were also reported. All tests were 2-sided, and P-values less than 0.05 were considered significant. Statistical analyses were conducted using R (version 4.2).

Bias

To minimize bias in this retrospective study, uniform eligibility criteria were applied to reduce selection bias, and complete-case analysis was used for patients with available survival and molecular data. Information bias was limited by extracting clinical, treatment, and outcome variables from standardized electronic health records using consistent RECIST (version 1.1) and CTCAE (version 5.0) definitions. Key variables (e.g., stage, histology, molecular alterations) were verified using pathology review or validated genomic reports to reduce misclassification. Confounding was addressed by incorporating clinically relevant covariates into multivariable Cox models.

Author contributions

H.N.T. and E.E.D. conceived the study and designed the analysis. H.N.T., L.K., and H.L. collected all clinical and molecular data. H.N.T. performed statistical analyses and drafted the manuscript. E.E.D., A.N., M.A., F.F., F.M.-B., S.C., E.S., and K.B.H.F. provided clinical oversight, interpretation, and critical revisions. All authors reviewed and approved the final manuscript.

Data availability

The datasets generated and/or analyzed during the present study are not publicly available due to institutional privacy policies, but are available from the corresponding author upon reasonable request and with appropriate institutional permissions.

Code availability

No custom code was used in this study.

Competing interests

H.N.T., H.L., K.B.H.F., L.K., and F.F. have no relationships to disclose. E.S. has a consulting/advisory role for D.E. Shaw Research; owns a patent for atezolizumab for treatment of alveolar soft part sarcoma; and participated on a data safety monitoring board or advisory board for FortVita and Mallinckrodt/Therakos. S.C. Speakers’ Bureau – Amgen, Astellas, AstraZeneca, Bristol Myers Squibb, Eisai, Genmab, Janssen, Merck KGaA, Merck, Novartis, Roche, Servier; Research Funding – Abbvie, Amgen, AstraZeneca, Boehringer Ingelheim, Bolt Biotherapeutics, Centessa Pharmaceuticals, Cytovation, Eisai, Exelixis, GlaxoSmithKline, Imcheck Therapeutics, Immunocore, Kumquat Biosciences, Marengo, Molecular Partners AG, Merck, OncoC4, Ose Immunotherapeutics, Owkin, Pheast, Pierre Fabre, Replimune, Roche, Sanofi Aventis, Seagen, Sotio A.S, Transgene, Trutino Biosciences, Consulting/Advisory Role – AccessTrial, Alderaan Biotechnology, Amgen; AstraZeneca, Aummune, Avacta, Bayer, Beigene, BioNTech, Celanese, Compugen, Domain Therapeutics, Ellipses Pharma, GE Health, Genmab, Guardant Health, Immunicom, Mariana Oncology, Mima Health, Nanobiotix, Nextcure, NetCancer, Oncovita, Pharma Mar, Pierre Fabre, Replimune, Seagen, Takeda, Tatum Bioscience, Tollys, UltraHuman8; Travel/Accommodations/Expenses – Amgen, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Merck, Ose Immunotherapeutics, Roche, Sotio; Stocks – Avacta. F.M.-B. Employment – MD Anderson Cancer Center; Honoraria – Dava Oncology; Consulting/Advisory Role – AstraZeneca, Becton Dickinson, Calibr, Cybrexa Therapeutics, Daiichi Sankyo/AstraZeneca, DAVA Pharmaceuticals, Debiopharm Group, EcoR1 Capital, eFFECTOR Therapeutics, Elevation Oncology, Exelixis, GT Aperion, Guardant Health, Harbinger Health, Incyte, Jazz Pharmaceuticals, LegoChem Biosciences, Loxo/Lilly, Menarini Group, Mersana, Molecular Templates, OnCusp Therapeutics, Protai, Ribometrix, Seagen, Tempus, Theratechnologies, Zentalis, Zymeworks; Research Funding – Aileron Therapeutics (Inst), AstraZeneca (Inst), Bayer (Inst), Boehringer Ingelheim (Inst), Calithera Biosciences (Inst), Curis (Inst), CytomX Therapeutics (Inst), Daiichi Sankyo (Inst), Debiopharm Group (Inst), eFFECTOR Therapeutics (Inst), Genentech (Inst), Guardant Health (Inst), Jazz Pharmaceuticals (Inst), Klus Pharma (Inst), Novartis (Inst), Pfizer (Inst), PUMA Biotechnology (Inst), Seagen (Inst), Taiho Pharmaceutical (Inst), Takeda Pharmaceutical (Inst), Zymeworks (Inst); Travel/Accommodations/Expenses – Cholangiocarcinoma Foundation, Dava Oncology, ESMO, EORTC. A.N. Employment – Baylor College of Medicine and Texas Children’s Hospital (I), MD Anderson Cancer Center; Honoraria – AKH Inc., American Society of Clinical Oncology, CME Outfitters, ESMO, Korean Society of Medical Oncology (KSMO), Lynx Group, Scripps Hospital, Society for Immunotherapy of Cancer; Consulting/Advisory Role – Abbvie, CytomX Therapeutics, Deka Biosciences, Genome & Company, Horizon Therapeutics (I), Immune-Onc Therapeutics, Janssen Biotech, Lynx Health, Merck Sharp & Dohme Corp, NGM Biopharmaceuticals, Nouscom, OncoSec, Pharming NV (I), PsiOxus Therapeutics, Servier, STCube Pharmaceuticals Inc., Takeda (I); Speakers’ Bureau – Alfaisal University (I); Research Funding – Amplimmune, Arcus Biosciences, ARMO BioSciences, Atterocor, Baxalta (I), BioNTech SE, Bristol-Myers Squibb, Calithera Biosciences, Chao Physician-Scientist Awards (I), CytomX Therapeutics, EMD Serono, Healios, Immune Deficiency Foundation (I), ImmuneOncia, Incyte, Jeffrey Modell Foundation (I), Karyopharm Therapeutics, Kymab, Lilly, MedImmune, Merck, Monopteros Therapeutics, NCI, NeoImmuneTech, Neon Therapeutics, Novartis, Pfizer, PsiOxus Therapeutics, Regeneron, Seven and Eight Biopharmaceuticals, Sotio, Surface Oncology, Texas Medical Center Digestive Diseases Center (I), TopAlliance BioSciences Inc; Travel/Accommodations/Expenses – ARMO BioSciences, NeoImmuneTech, NGM Biopharmaceuticals. M.A. Research Funding – Genentech (Inst), Nektar Therapeutics (Inst), Merck (Inst), GlaxoSmithKline (Inst), Novartis (Inst), Jounce Therapeutics (Inst), Bristol Myers Squibb (Inst), Eli Lilly (Inst), Adaptimmune (Inst), Shattuck Lab (Inst), Gilead (Inst), BioNTech (Inst), Henlius (Inst), Vivace Therapeutics (Inst), Verismo Therapeutics (Inst), Pfizer (Inst); Consulting/Advisory Role – GlaxoSmithKline, Shattuck Lab, Bristol Myers Squibb, AstraZeneca, Insight; Honoraria – AstraZeneca, Nektar Therapeutics, Society for Immunotherapy of Cancer (SITC); Participation on a Data Safety Monitoring Board or Advisory Board – Nanobiotix–MDA Alliance, Henlius. E.E.D. Consulting/Advisory Role – BOLT Biotherapeutics, Fate Therapeutics, Mersana, PMV Pharma, Summit Therapeutics; Speakers’ Bureau – PMV Pharma; Research Funding – A2A Pharmaceuticals (Inst), Aileron Therapeutics (Inst), Amgen (Inst), Aprea Therapeutics (Inst), Astex Pharmaceuticals (Inst), AstraZeneca (Inst), Bayer (Inst), Bellicum Pharmaceuticals (Inst), Boehringer Ingelheim (Inst), BOLT Biotherapeutics (Inst), Compugen (Inst), Dragonfly Therapeutics (Inst), Fate Therapeutics (Inst), Genentech (Inst), Immunocore (Inst), Immunomedics/Gilead (Inst), Mereo BioPharma 5 (Inst), Mersana (Inst), ModeX Therapeutics (Inst), Pfizer (Inst), PMV Pharma (Inst), Poseida (Inst), Rain Therapeutics (Inst), Sanofi (Inst), Sotio (Inst), Triumvira Immunologics (Inst), Volastra Therapeutics (Inst); Travel/Accommodations/Expenses – Boehringer Ingelheim, Rain Oncology, Triumvira Immunologics.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41698-026-01516-9.