A comprehensive analysis and visualization of immune-related adverse events in lung cancer immunotherapy a bibliometric study

Abstract

Background

Immune-related adverse events (irAEs) significantly impact the therapeutic efficacy of immune checkpoint inhibitors (ICIs) in lung cancer. However, a comprehensive understanding of the global research landscape and its evolution regarding irAEs is still lacking.

Methods

We conducted a bibliometric analysis of 6787 publications from the Web of Science Core Collection (2015–2024) via CiteSpace and the R-bibliometrix package. The analyses included cocitation, coauthorship, and keyword burst detection to map collaborative networks, knowledge structure, and research trends.

Results

Research output demonstrated a remarkable compound annual growth rate of 30.4%. INSERM (France) and US institutions appeared as central hubs in collaborative networks, whereas China—despite its high publication volume—showed more limited international integration, suggesting a productivity‑influence mismatch. The knowledge base comprises three clusters: clinical evidence, translational research, and fundamental immunology. Research themes evolved from initial tumor-agnostic safety reports to a focus on organ-specific irAEs and precision toxicology.

Conclusion

This study delineates the rapidly evolving landscape of irAE research in lung cancer immunotherapy, highlighting key collaborative patterns, knowledge structures, and thematic shifts. The findings provide a foundational perspective for guiding future research directions.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05354-w.

Introduction

Lung cancer remains the leading cause of cancer-related mortality globally [1]. Immunotherapy, particularly immune checkpoint inhibitors (ICIs) that target molecules such as programmed cell death protein 1 (PD-1), programmed death-ligand 1 (PD-L1), and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), has emerged as a cornerstone of lung cancer treatment, significantly improving overall survival rates in subsets of patients, including those with advanced non-small cell lung cancer (NSCLC) [2]. However, ICIs are associated with a unique spectrum of adverse events known as immune-related adverse events (irAEs), which can affect multiple organ systems (e.g., the lungs [immune-related pneumonitis], skin [rash, pruritus], gastrointestinal tract [colitis], endocrine glands [thyroid dysfunction], liver [hepatitis], and nervous system [neuropathy]) [3]. The mechanisms, clinical manifestations, and optimal management strategies for irAEs exhibit substantial heterogeneity. This heterogeneity impacts patients’ quality of life and, in severe cases, can lead to treatment discontinuation or death [4]. For example, Brahmer et al. [5] reported high-grade irAEs (grades 3–4) in approximately 10–15% of patients receiving PD-1 inhibitors, although the incidence varies by organ system and ICI class, a variability not fully explained by current mechanistic models [5–7].

While the overall incidence of irAEs is generally lower than that of toxicities associated with traditional cytotoxic chemotherapy, irAEs are characterized by unpredictable onset. Certain irAEs, such as immune-mediated pneumonitis and myocarditis, can develop abruptly, progress rapidly, and have high fatality rates, often requiring multidisciplinary management [8]. Notably, the incidence of immune-mediated pneumonitis is significantly greater in lung cancer patients than in patients with other solid tumors, potentially because of preexisting lung pathology or lung tissue immunogenicity, which warrants specific clinical attention. Currently, the pathophysiological mechanisms driving irAEs are not fully understood; for example, the role of tissue-resident memory T cells in organ-specific toxicity remains under investigation [9]. Additionally, reliable predictive biomarkers (e.g., genetic variants and serum cytokines) and individualized management strategies are lacking. This poses major clinical challenges [10, 11]. Existing studies suggest that the occurrence of irAEs is associated with ICI type (e.g., PD-1 vs. CTLA-4 inhibitors) and patient-specific factors, including biomarker status (e.g., PD-L1 expression), genetic background (e.g., HLA haplotypes), and microbiome composition. For example, Rizvi et al. [11] reported a potential link between tumor mutational burden (TMB) and irAE risk, although subsequent studies reported conflicting results, highlighting the need for larger, multicohort validation [12]. Notably, some studies indicate a potential correlation between irAE occurrence and treatment efficacy, where certain irAEs (e.g., mild rash) may predict better therapeutic responses. However, this correlation is not universal, and the biological basis of this association—whether reflecting enhanced antitumor immunity or off-target immune activation—remains unclear [13].

As immunotherapy application in lung cancer expands—from advanced-line treatment to first-line, neoadjuvant, and adjuvant settings—research on irAEs has grown exponentially. This surge includes clinical trials evaluating ICI safety, real-world data on irAE incidence in diverse populations, and translational studies investigating toxicity mechanisms. The volume of outputs necessitates systematic synthesis to clarify the field’s knowledge foundation, identify current research hotspots, and trace evolutionary trends. Bibliometric methods combined with visualization techniques offer an objective, quantitative approach to revealing disciplinary dynamics, mapping core research forces and collaborative networks, and providing strategic insights for scholars. Such analyses may also help identify understudied areas—for example, irAE management in elderly or comorbid patients—that warrant increased research attention.

This study employs bibliometric methods to conduct a multilevel analysis of global research literature on irAEs in lung cancer immunotherapy (2015–2024) [14, 15]. The analysis includes collaborative patterns across countries/regions; the contributions and influence of authors and institutions; and the distribution and impact of journals. Additionally, it examines keyword co-occurrence and clustering, research theme evolution, and the identification of emerging research frontiers. Several prior bibliometric studies have explored irAEs in immunotherapy, including Jiang et al. [14], Gong et al. [15], and Zheng et al. [16]. However, these studies either focused on broader irAE contexts without lung-specific analysis, covered shorter timeframes, or employed single analytical tools. Our study provides a comprehensive update covering 2015–2024, integrates dual tools (CiteSpace and R-bibliometrix) for cross-validation, and offers an in-depth examination of collaborative networks, knowledge clusters, and thematic transitions, thereby offering novel insights into the evolving landscape of irAE research specifically in lung cancer immunotherapy. The primary aim is to elucidate the field’s knowledge structure and research trends, offering evidence-based insights and novel theoretical and clinical perspectives for optimizing lung cancer immunotherapy strategies. A deeper understanding of irAEs through this mapping is expected to enhance patients’ quality of life and treatment efficacy—although translating these insights into clinical practice will depend on integrating bibliometric findings with empirical clinical research [16].

Methods

Data sources and search strategy

Data were retrieved from the Web of Science Core Collection, specifically the Science Citation Index Expanded (SCI-E) database, which was selected for its comprehensive coverage of high-impact, peer-reviewed biomedical literature. The search was conducted on September 1, 2025, covering publications from January 1, 2015, to December 31, 2024. A stepwise thematic search strategy ensured dataset specificity and relevance: first, a core dataset focused on “lung cancer” and “immunotherapy” was constructed, followed by screening with terms related to “adverse reactions” to isolate irAE studies.

Specific search queries: Query 1 (#1: Lung cancer): TS=((lung OR pulmonary OR “non-small cell” OR NSCLC) NEAR/3 (cancer* OR carcinoma* OR tumor* OR tumour* OR neoplas* OR adenocarcinoma*)).

Query 2 (#2: Immunotherapy): TS=(“immune checkpoint inhibitor*” OR immunotherapy OR “pd-1” OR “pd-l1” OR “ctla-4” OR “programmed death 1” OR “programmed death ligand 1” OR ipilimumab OR nivolumab OR pembrolizumab OR atezolizumab OR durvalumab OR cemiplimab OR sintilimab OR tislelizumab).

Query 3 (#3: Adverse events): TS=((adverse event* OR adverse reaction* OR toxicity OR toxicities OR side effect* OR “safety profile” OR immune-related OR irAE* OR irEC* OR irC* OR irP* OR irD* OR irColitis OR irHepatitis OR irNephritis OR irMyocarditis OR irPneumonitis OR irRash OR irArthritis OR “immune-mediated” OR “treatment-related” OR “treatment-emergent”)).

Final Query (#4): #1 AND #2 AND #3.

In Query 1, both “tumor” and “tumour” are included to capture both US and UK English spelling variants, ensuring comprehensive retrieval of relevant literature.The document types were restricted to “Articles” and “Reviews” published in English. This restriction was applied as part of the initial search query in the Web of Science interface using the built-in document type filter, ensuring that all retrieved records conformed to these types from the outset, rather than being applied as a post-retrieval filter. The initial search yielded 7,320 publications.

The NEAR/3 operator in Query 1 was selected to balance retrieval sensitivity and specificity by capturing closely related lung-cancer expressions while minimizing unrelated noise from broad oncology terms.

Although Web of Science is widely used in bibliometric research for its high-quality indexing and consistent citation data, we acknowledge that its exclusive use may underrepresent non-English publications and grey literature. This potential bias should be considered when interpreting the geographic and institutional patterns reported here (see also Sect.  4.8).

Manual screening and eligibility criteria

The study selection process followed a two-stage screening approach (title/abstract followed by full-text review) conducted independently by two researchers, with discrepancies resolved through consensus.

Inter-rater reliability for independent screening was quantified using Cohen’s kappa coefficient before consensus adjudication. The obtained kappa value of 0.89 indicated excellent agreement between the two independent reviewers.

The inclusion criterion was as follows: study type: original research articles or reviews. Population: Human subjects diagnosed with lung cancer (any stage or histology).

Intervention: Treatment with immune checkpoint inhibitors (as monotherapy or in combination). Outcome: Explicit reporting of data on the incidence, spectrum, management, or outcomes of immune-related adverse events (irAEs).

The exclusion criterion was studies not involving human subjects (e.g., in vitro or animal studies). Non-English publications. Studies that did not focus primarily on lung cancer immunotherapy. Studies that did not report specific irAE data (e.g., focusing solely on efficacy without a safety profile).

Following deduplication (no duplicates were identified as only one database was searched), rigorous manual screening was performed. All exclusions, including the removal of non-English publications, were conducted manually by the two independent reviewers during title/abstract and full-text screening; no automated exclusion tools or language filters were used. A total of 533 records were excluded. The reasons for exclusion included nonhuman studies (n = 394), non-English manuscripts (n = 67), and studies failing to meet the inclusion criteria regarding irAE reporting or lung cancer focus (n = 72). The relatively low number of exclusions for “lack of irAE focus” (n = 72) reflects the effectiveness of our combined search strategy: by requiring the intersection of all three thematic queries (lung cancer AND immunotherapy AND adverse events), the initial retrieval was already highly specific to irAE-related research. The 72 excluded papers represent cases where, despite meeting the search criteria, the study did not primarily focus on irAEs (e.g., mentioning adverse events only incidentally in the context of efficacy reporting). A total of 6787 publications were included in the analysis (Fig. 1).

Fig. 1.

Study selection flowchart following the PRISMA 2020 guidelines for the systematic analysis of immune-related adverse events (irAEs) associated with lung cancer immunotherapy. No duplicate removal across databases was performed, as only one database was searched. All exclusions were performed by human reviewers; no automation tool was used

Analytical tools and parameter settings

Two complementary bibliometric analysis tools were employed: CiteSpace (6.2. R3, 64-bit) [16] and the R-bibliometrix package (5.1.1) [18]. CiteSpace was used for document cocitation analysis, keyword burst detection, and generation of timeline and time zone views. For author co-citation analysis, CiteSpace includes all authors (first, last, and co-authors) from each cited reference by default, counting co-citations when any two authors appear together in the reference lists of citing papers. Core settings included a 1-year time slice and pathfinder/sliced-network pruning. The g-index parameter was tuned by network type to optimize readability and structural stability: most analyses used k = 25, while Fig. 5 (journal co-citation) used k = 5 and Fig. 6 (author co-citation) used k = 8.

Fig. 5.

Journal co-citation network analysis in immune-related adverse events (irAEs) research related to lung cancer immunotherapy (2015–2024). Network visualization showing the intellectual structure of the research field through journal co-citation analysis. Nodes represent journals, with node size proportional to the number of times the journal was cited. Edge thickness indicates the strength of co-citation relationships between journals. Colors represent different clusters of related journals identified through modularity-based community detection. Labels show the most influential journals with high centrality measures. The network was constructed using CiteSpace with time slicing (2015–2024), node selection criteria (top 50 per slice), and pruning algorithms to reduce network complexity while preserving important structural information

Fig. 6.

Author cocitation network for irAE research related to lung cancer immunotherapy (2015–2024). Author co-citation analysis revealing the intellectual structure and key contributors to the field. Nodes represent authors, sized by their citation frequency and research influence. Edge connections indicate co-citation relationships, with thickness proportional to the frequency of joint citations. Color clusters represent different research schools or methodological approaches identified through cluster analysis. Author names are displayed for researchers with high centrality scores and significant contributions to the field. Network parameters were optimized to highlight influential authors while maintaining visual clarity and interpretability

R-bibliometrix was used to extract core bibliometric indicators (annual publication output, country/regional productivity, h index, collaboration degree) and generate scientific maps (e.g., country collaboration networks). Built-in functions for data cleaning (e.g., institutional name standardization) minimized network analysis bias.Detailed search strategies, parameter settings for each figure, and screening data are available in the supplementary file.

Indicator definitions

Key metrics used for interpretation: Burst strength: This metric identifies keywords or references that have experienced a significant, rapid increase in citation frequency over a specific period. The threshold of ≥ 10 was selected on the basis of methodological conventions in the cited software’s literature [40] and was empirically validated as effective for isolating terms with substantial and sustained scholarly impact within our dataset. This approach prioritizes the most prominent research trends; however, it is acknowledged that this may exclude emerging topics with currently lower—but potentially growing—burst strength.

Betweenness Centrality: This metric quantifies a node’s role as a bridge or intermediary between different communities within the network. Nodes with a centrality value ≥ 0.1 are typically classified as significant intermediaries, although this interpretation may be influenced by overall network density and structure.

Modularity (Q) and mean silhouette (S): These metrics evaluate the effectiveness of the network’s clustering. A Q value > 0.3 indicates that the network possesses a significant cluster structure. An S value > 0.5 suggests reasonable internal consistency within clusters, whereas S > 0.7 indicates highly convincing and distinct clustering. The interpretation of these values is contextualized by the size and complexity of the network under study. All indicator definitions and calculations follow the specifications implemented in the CiteSpace software and detailed by [16, 17].

Sensitivity analysis of key parameters

To assess the robustness of our main findings, we performed sensitivity tests for two critical parameters: the burst detection γ value (set at 0.3) and the g‑index thresholds (k = 25, with k = 5 and k = 8 for specific figures). Varying γ from 0.2 to 0.4 produced only minor changes in the list of burst keywords (variation < 5%), and the three‑stage thematic evolution (Foundation, Validation, Optimization) remained consistent. Similarly, adjusting the g‑index by ± 2 for the author co‑citation network did not alter the identity of the top‑ranked authors or the overall cluster structure. These checks support the stability of our main conclusions under reasonable parameter variations.

Results

Global publication trends (2015–2024)

The global literature on irAEs in lung cancer immunotherapy has revealed a marked, sustained upward trend (Fig. 2). The number of annual publications increased from 106 in 2015 to a peak of 1,167 in 2022 (CAGR = 30.4%). This growth aligns with successive worldwide regulatory approvals of ICIs (e.g., FDA approval of pembrolizumab for first-line NSCLC in 2016) and their widespread clinical adoption [19, 20]. Research themes have diversified, shifting from early mechanistic investigations to phase I–III clinical trials evaluating ICI safety. More recently, real-world evidence studies have examined the incidence and management of irAEs in diverse populations, such as elderly and comorbid patients [21, 22]. As of September 30, 2024, the field had accumulated 352,410 total citations, with a mean of 51.35 citations per paper and an h-index of 246, underscoring significant research activity and scientific impact, although citation patterns can be influenced by journal impact factors and publication timing [21–23].

Fig. 2.

Annual Publications and Times Cited for irAEs in Lung Cancer Immunotherapy (2015–2024). A Annual number of publications showing the temporal distribution of research output. B Cumulative citation counts per year reflecting research impact and academic influence. Data were extracted from Web of Science Core Collection and analyzed using bibliometric methods. The y-axis represents the number of publications (A) and cumulative citations (B), respectively. Error bars represent 95% confidence intervals. The temporal scope was limited to 2015–2024 to focus on contemporary research trends. Statistical significance was assessed using linear regression analysis (p < 0.05)

National and institutional collaboration patterns (2015–2024)

National collaboration network

The country-level coauthorship network (derived from 4,812 records) consists of 105 nodes (countries) and 124 edges (collaborative relationships), with a density of 0.0227 (Fig. 3). The network exhibits a core–periphery topology: the United States (US) has the highest degree centrality (0.43) and maintains direct collaborative ties with 48 countries, functioning as the primary hub for global knowledge diffusion. China ranks second in degree centrality (0.28), with 36 collaborative links. Its strongest bilateral collaboration is with the US (92 coauthored papers), followed by Japan (23) and Australia (14). Within Europe, Germany, France, the United Kingdom (UK), and Italy form a densely connected subcluster, with an internal density of 0.31—significantly higher than the global mean (p < 0.01). Emerging Asian economies (South Korea, Singapore, India) serve as bridging nodes between East Asian and European/North American clusters. In contrast, Latin American, African, and Eastern European countries remain at the periphery with limited collaborative activity. This pattern illustrates a Matthew effect, whereby core nodes attract additional collaborations, reinforcing their centrality [24]. A geographical map with latitude and longitude coordinates (Fig. 4) confirms three primary collaboration hotspots: the North American corridor (US–Canada), the European backbone (Germany–France–UK–Italy), and the East Asian triangle (China–Japan–South Korea). These hotspots are interconnected by trans-Pacific and trans-Atlantic links, whereas collaboration activity remains limited in South America, Africa, and Central Asia.

Fig. 3.

Country Collaboration Network in irAE Research Related to Lung Cancer Immunotherapy (2015–2024). Network visualization displaying international research collaborations among countries. Node size represents the number of publications from each country, with larger nodes indicating higher research output. Edge thickness represents the strength of collaboration, determined by the number of co-authored publications between countries. Colors indicate different research clusters identified through community detection algorithms. The network was constructed using CiteSpace software with parameters set to show countries with minimum document counts ≥ 10. Layout was generated using the Kamada-Kawai algorithm to optimize node positioning and minimize edge crossings

Fig. 4.

Country Collaboration Map of irAE Research Related to Lung Cancer Immunotherapy. World map visualization showing research collaboration patterns across different geographical regions. Country colors represent the intensity of research collaboration, with darker colors indicating higher collaboration frequencies. Line connections between countries represent collaborative research relationships, with line thickness proportional to the number of joint publications. Data were extracted from Web of Science Core Collection and visualized using bibliometric mapping techniques. The map was generated using standard geographic projections to ensure accurate spatial representation of research collaboration patterns

Institutional collaboration network

The institutional network comprised 353 organizations (mean degree = 5.4, clustering coefficient = 0.42), indicating small-world properties. While Chinese institutions ranked highly in output—Chinese Academy of Medical Sciences (229 articles) and Peking Union Medical College (170 articles)—their betweenness centrality was modest (0.12 and 0.10, respectively), indicating limited influence on cross-institutional knowledge flow. INSERM (France) led in betweenness centrality (0.68) despite a slightly lower output (240 articles). It connected 178 institutions across 42 countries and formed a transcontinental triangle with the UT MD Anderson Cancer Center and the Chinese Academy of Medical Sciences. UT MD Anderson (249 articles; centrality = 0.33) and the University of California System (213 articles; centrality = 0.31) formed the dual core among U.S. institutions. Overall, Chinese institutions were semiperipheral, channeling only 17.6% of the global information flow relative to INSERM—a gap likely due to their focus on domestic rather than international multicenter studies.

Temporal evolution of collaboration

Burst detection analysis (γ = 1.0, minimum duration: 2 years) identified China (burst strength: 19.7; 2018–2021) and South Korea (burst strength: 8.3; 2016–2019) as the most dynamically growing national nodes. At the institutional level, the Chinese Academy of Medical Sciences (burst strength: 11.2; 2020–2024) and the University of Sydney (burst strength: 9.8; 2017–2020) emerged as the strongest rising collaborators. These trends suggest a gradual shift from a U.S.–Europe bipolar structure toward a U.S.–Europe–Asia–Pacific tripolar configuration, whereas the core U.S./European influence remains intact. The irAE collaboration network is characterized by low overall density but high clustering within regional and thematic subgroups. The U.S. maintains structural dominance in degree centrality, whereas INSERM operates as the global central hub for cross-cluster collaboration. China, despite its high output, remains dependent on bilateral collaborations with core nodes. Transitioning to a hub-and-spoke architecture through participation and leadership at multiple centers, intercontinental consortia are critical for converting quantity into network influence; however, the feasibility of this transition may be constrained by differences in regulatory frameworks and data-sharing policies, which are recognized as significant barriers in international clinical and translational research [22].

Journal cocitation network: knowledge base and evolution

Figure 5 visualizes the journal cocitation landscape, comprising 538 nodes (journals) and 3,449 weighted edges (cocitation relationships), with a density of 0.0024.The network exhibited high modularity (Q = 0.7318) and strong internal coherence (weighted mean silhouette = 0.8762), indicating distinct, thematically consistent clusters. Three dominant knowledge clusters were identified:

Clinical-Evidence Core (Red Cluster, 34% of Citations): This cluster is anchored by high-impact clinical journals such as the New England Journal of Medicine (NEJM), The Lancet, JAMA, and Lancet Oncology, with a mean of over 180 citations per node. These journals published pivotal phase III trials (e.g., KEYNOTE-024 [23] and CheckMate-227 [24, 25]) that defined irAE grading criteria and discontinuation thresholds. This cluster represented the segment with the highest citation burst (burst strength = 18.7, 2018–2021) and serves as the backbone for clinical guidelines, although the generalizability of findings to real-world populations remains under discussion.

Translational-Methodology Bridge (Green Cluster, 29% of Citations): Journals such as the Journal for ImmunoTherapy of Cancer, Clinical Cancer Research, and Nature Reviews Clinical Oncology function as methodological gatekeepers, with centrality scores ranging from 0.38 to 0.42. They focus on translating mechanistic insights into clinical tools, such as predictive biomarkers. Since 2020, emerging journals (e.g., Signal Transduction and Targeted Therapy, Molecular Cancer; burst strength = 9.4) have disseminated studies utilizing single-cell RNA sequencing and multiomic profiling to develop irAE predictive models. However, clinical validation of these models is critically lacking [26].

Fundamental Immunology and Molecular Mechanisms (Blue Cluster, 37% of Citations): This cluster includes journals such as Nature Reviews Drug Discovery, Immunity, and Cell Reports, which constitute the mechanistic layer. It exhibited the fastest citation growth, with a compound annual growth rate (CAGR) of 21% compared with 13% for the clinical cluster between 2019 and 2024. This growth reflects a shift from descriptive reporting to precision interrogation of the immune microenvironment and germline–somatic interactions. Open-access journals (e.g., Frontiers in Immunology, Cells) entered the top five bursting journals after 2022, highlighting their rising role in disseminating mechanistic evidence [27].

Overall, irAE research has evolved from a clinical-case-driven narrative into a tripartite ecosystem. The high modularity and low density of the network suggest knowledge expansion along specialized lines with limited cross-cluster integration. Boundary-spanning journals, such as the Journal of Thoracic Oncology and Annals of Oncology, are critical for synthesizing mechanistic insights into clinical guidelines. For Chinese researchers, strategically targeting high-centrality, translation-oriented journals is important for publishing studies on irAE prediction and intervention [28].

Author cocitation network and authority identification

For author co-citation analysis, we included all authors (first, last, and co-authors) as per CiteSpace default settings.

Network robustness and clustering quality

Applying CiteSpace (g-index = 8, pathfinder/sliced network pruning) [6] yielded an author cocitation network comprising 538 nodes (authors) and 349 edges (cocitation relationships), with a density of 0.0024. High modularity (Q = 0.7318) and coherence (silhouette = 0.8762) confirmed a highly modular, thematically consistent configuration (Fig. 6).

Intellectual gatekeepers: centrality and productivity

The top-betweenness “core triangle” included the following authors:

1.  Brahmer JR (Johns Hopkins): Centrality=0.41; citations=1887. Focus: Early PD-1 inhibitor development and irAE characterization (especially pneumonitis).

2.  Herbst RS (UT MD Anderson): Centrality=0.38; citations=1654. Focus: First-line ICI regimens and irAE risk biomarkers.

3. Paz-Ares L (CNIO): Centrality=0.35; citations=1421. Focus: Phase III ICI combination trials and irAE management guidelines.

These authors collectively accounted for a substantial portion of the inter‑cluster knowledge flow (as measured by betweenness centrality), although this metric captures only formal co‑citation relationships and may not fully reflect all forms of intellectual influence. The highest-ranked Chinese scholar, Wu YL, ranked 39th (centrality = 0.07), indicating that China lacks intellectual gatekeepers who significantly influence the global network. This may be due to a focus on domestic questions or limited participation in international trials shaping the core discourse.

Knowledge clusters and thematic chronology

Log-likelihood ratio clustering revealed seven groups, with three dominant streams:

Cluster #0—Clinical Immuno-Oncology Foundation (S = 0.902): Key authors include Brahmer JR, Herbst RS, Gandhi L, Paz-Ares L, and Garassino MC. The signature terms used were “PD-1 inhibitor,” “phase III trial,” and “first-line NSCLC.” Burst strength = 34.7 (2016–2018), aligning with first-line ICI approvals. This cluster established ICI monotherapy and chemoimmunotherapy standards; however, questions remain regarding optimization strategies for patients at high risk of irAEs.

Cluster #1—Global Cancer Burden and Theoretical Framework (S = 0.891): Key authors include Bray F, Siegel RL, Hanahan D, and Ferlay J. Signature terms: “global cancer statistics,” “hallmarks of cancer.” Compound Annual Growth Rate (CAGR) = 11.4%. This cluster provides a population-level context for irAE epidemiology and theoretical frameworks; however, its application to predict individual irAE risk remains limited.

Cluster #2—Precision Medicine and Resistance (S = 0.884): Key authors include Hellmann MD, Forde PM, Arbor KC, and Rizvi NA. The signature terms used were “TMB,” “ctDNA,” and “primary/secondary resistance.” Burst strength = 28.3 (2019–2024). This cluster focuses on neoadjuvant immunotherapy, next-generation ICIs (LAG-3, TIGIT), and liquid biopsy-guided irAE monitoring. The clinical utility of TMB and ctDNA for irAE prediction requires further validation.

Generate turnover

Fifteen authors exhibited significant citation bursts, reflecting three evolutionary waves:

Wave 1 (2015–2017): Dose-Finding Pioneers—Brahmer JR (burst = 34.7), Topalian SL (burst = 26.9). Focus: Initial immune checkpoint inhibitor (ICI) clinical development, dose escalation, and early irAE profiles.

Wave 2 (2018–2020): Combination strategies—Gandhi L (burst = 22.4) and Paz-Ares L (burst = 20.1). Focus: ICI combination safety (e.g., PD-1 and CTLA-4 inhibitors), shaping clinical guidelines.

Wave 3 (2021–2024): Precision Medicine and Resistance Mechanisms—Forde PM (burst = 19.8), Arbor KC (burst = 18.6). Focus: Precision irAE prediction using multiomic biomarkers and irAE management in resistant patients. This progression reflects the field’s shift from demonstrating proof-of-efficacy toward precision medicine and understanding relapse biology, although integration into clinical practice remains ongoing.

Chinese scholars’ performance

Eighteen Chinese researchers have ranked among the top 100 most-cited authors; however, none have exhibited centrality greater than 0.1 or significant citation bursts. They have formed dense domestic subclusters with few trans-Pacific links. To achieve greater influence, it is necessary to lead multicenter phase III trials addressing unmet needs (e.g., irAE management in elderly patients) and to publish practice-changing evidence [29]. This ability depends on overcoming barriers such as access to international networks and regulatory alignment [30].

Document cocitation burst analysis: milestones and turning points

Dataset and parameters

Two time periods were analyzed: a long-term period (2012–2024) to examine foundational development and a short-term period (2018–2024) to capture recent trends. For the parameter settings, a parameter γ = 0.3 was used to balance the burst detection sensitivity and specificity. Additionally, a minimum burst duration of 2 years was set to exclude transient citations. Fifty-six burst references were identified, and 31 references with burst strengths ≥ 10 were analyzed (Table 1).

Table 1.

Highly cited references with the strongest citation bursts in irAE research (2015–2024)

Rank	First author	Year	Journal	Burst strength	Burst interval
A. Historical burst (2015–2021)
1	Topalian SL	2012	N Engl J Med	113.96	2015–2017
2	Robert C	2015	N Engl J Med	72.2	2015–2019
3	Pardoll DM	2012	Nat Rev Cancer	68.47	2015–2017
4	Rizvi NA	2015	Lancet Oncol	65.6	2015–2018
5	Rizvi NA	2015	Science	58.78	2015–2020
6	Herbst RS	2014	Nature	57.99	2015–2019
7	Weber JS	2015	Lancet Oncol	55.41	2015–2019
8	Brahmer JR	2012	N Engl J Med	53.16	2015–2017
9	Hamid O	2013	N Engl J Med	52.79	2015–2018
10	Wolchok JD	2013	N Engl J Med	51.83	2015–2018
B. Current burst (2022–2024)
1	Sung H	2021	CA Cancer J Clin	120.08	2022–2024
2	Paz-Ares L	2021	Lancet Oncol	61.66	2022–2024
3	Herbst RS	2020	N Engl J Med	53.07	2022–2024
4	Reck M	2021	J Clin Oncol	41.32	2022–2024
5	Felip E	2021	Lancet	39.22	2022–2024
6	Sezer A	2021	Lancet	39.18	2022–2024
7	Zhou CC	2021	Lancet Resp Med	32.31	2022–2024
8	Liu SV	2021	J Clin Oncol	30.21	2022–2024
9	Siegel RL	2022	CA Cancer J Clin	29.31	2022–2024
10	West H	2019	Lancet Oncol	28.59	2022–2024

Sensitivity checks were performed for key parameters (γ and g‑index) as detailed in Methods 2.5; the main findings described in Sects.  3.5.2, 3.5.3, 3.5.4 and 3.5.5 were robust to these variations.

Burst magnitude and duration

Between 2012 and 2024, the strongest citation burst was observed for the study by Topalian SL et al. (N Engl J Med, 2012), which demonstrated the first clinical proof-of-concept for PD-1 blockade [31]. This burst had an intensity of 231.6 and occurred from 2013 to 2017. The second strongest burst was associated with the KEYNOTE-024 trial by Reck M et al. (N Engl J Med, 2016), which showed the superiority of pembrolizumab in PD-L1-positive NSCLC [26]; this burst had an intensity of 130.6 and spanned from 2017 to 2021.

From 2018 to 2024, the highest citation burst was recorded for the global cancer statistics paper by Sung H et al. (CA Cancer J Clin, 2021), with a burst intensity of 120.1 occurring between 2022 and 2024 [32]. This work emphasized the importance of the population-level epidemiology of irAEs [33]. Additionally, recent publications by Paz-Ares L and Herbst RS have exhibited sustained citation bursts, reflecting ongoing interest in this area of research (Fig. 7).

Fig. 7.

Cocitation clustering of highly cited papers in irAE research related to lung cancer immunotherapy (2015–2024). Document co-citation analysis revealing the intellectual structure and research fronts in the field. Each node represents a highly cited publication, with node size proportional to citation frequency. Edge connections indicate co-citation relationships between papers. Colors represent different thematic clusters identified through bibliographic coupling and cluster analysis. Cluster labels were generated using log-likelihood ratio (LLR) and tf-idf algorithms to capture the main research themes. The visualization shows the evolution of research topics and the emergence of new research directions over the study period. Clusters were named based on log-likelihood ratio testing (p < 0.05) of the most frequent keywords extracted from the titles and abstracts of the citing articles within each cluster

Three-stage evolutionary trajectory

The timeline visualization clustered bursting references into three successive waves:

Foundation (2012–2017): Studies focused on mechanisms (median burst = 53.2, mean duration = 4.3 years). Key papers include Topalian SL (PD-1 inhibition) [5], Brahmer JR (ipilimumab irAEs) [34], and Pardoll DM (checkpoint biology) [35]. This stage involves the initial immune checkpoint inhibitor (ICI) dose, safety, and efficacy; however, the understanding of the mechanism remains limited.

Validation (2018–2021): Phase III randomized controlled trials (RCTs) formed the core of this period, with bursts often exceeding 80. The key trials included KEYNOTE-024 [36], CheckMate-026 [37], and IMPower150 [38]. This stage introduced concepts such as programmed death-ligand 1 (PD-L1) as a predictive biomarker and established irAE grading criteria. However, the clinical generalizability of these findings was limited due to the restrictive exclusion criteria used in these trials.

Optimization (2022–2024): Focus shifted to resistance biology and precision strategies. Papers by Herbst RS [39] and Paz-Ares L [40] (each with bursts ≥ 53 and durations ≥ 3 years) focused on long-term survival, novel target combinations (e.g., PD-1 and TIGIT), irAE subphenotyping, and multiomics data integration for prediction. Clinical implementation remains hindered by the lack of large validation cohorts (Fig. 8).

Fig. 8.

Timeline view of reference cocitation clusters in irAE research related to lung cancer immunotherapy (2015–2024). Temporal visualization showing the evolution of research themes and the emergence of new research fronts over time. Each horizontal line represents a research cluster, with the width proportional to the number of papers in that cluster. Timeline bars show the temporal distribution of cluster emergence and development. Colors indicate different thematic clusters, with newer clusters appearing in later time periods. The timeline view helps identify research frontiers and predict future research directions based on cluster dynamics and inter-cluster relationships

Core authors and institutional hubs

Seven authors contributed two or more papers with significant citation bursts (burst strength ≥ 50). Paz-Ares L (4 papers) and Herbst RS (5 papers) have contributed to research on ICI combinations, neoadjuvant therapy, and irAE management. Reck M linked European and US trials, including KEYNOTE-024 [36]. Yale University, UT MD Anderson, and Hospital 12 de Octubre led in producing these high-impact papers, with over 68% involving intercontinental coauthorship.

Journal and disciplinary landscape

Seventeen journals accounted for all references exhibiting citation bursts; among these, four journals (NEJM, Lancet Oncology, Journal of Clinical Oncology, and Nature Reviews Clinical Oncology) contributed 72.3% of the burst references. The interdisciplinarity index—which measures the overlap of immunology and oncology citations—increased from 41% in 2012 to 78.6% in 2024, confirming immuno-oncology as an independent subspecialty. Future breakthroughs will likely require multidisciplinary collaboration.

In summary, document cocitation burst analysis reveals a paradigm shift from “mechanistic proof” to “precision optimization,” which aligns with regulatory and clinical advancements. Future research hotspots include reversing ICI resistance, developing early irAE intervention models, and advancing multiomics-guided individualized immunotherapy. Progress in this field depends on access to large cohorts and effective data sharing.

Keyword co-occurrence bursts & thematic evolution (2015–2024)

Cluster topology

A keyword co-occurrence network (353 terms, 393 links, density = 0.0063) exhibited high modularity (Q = 0.861) and coherence (silhouette = 0.9569), indicating that the network consists of tightly knit, distinct groups (Fig. 9).The network clustering was constructed using log-likelihood ratio (LLR) testing, which identifies statistically significant keyword associations by comparing observed co-occurrence frequencies against expected frequencies under the null hypothesis of independence. This method extracts the most representative keywords for each cluster, which are then used to assign cluster labels. The clusters themselves were identified through modularity optimization, an algorithm that maximizes the density of connections within clusters while minimizing connections between clusters. The modularity Q value of 0.861 indicates a strong and reliable cluster structure. Log-likelihood ratio clustering resolved 17 communities, with the nine largest communities displayed in Fig. 10.

Fig. 9.

Keyword clustering network of irAE research in lung cancer immunotherapy (2015–2024). Keyword co-occurrence analysis revealing the main research themes and topics in the field. Nodes represent research keywords, sized by their frequency of occurrence in the literature. Edge connections indicate co-occurrence relationships between keywords. Colors represent different thematic clusters of related keywords identified through community detection algorithms. Cluster labels were automatically generated using TF-IDF weighting and statistical significance testing. The network visualization helps identify the core concepts, emerging topics, and knowledge structure of irAEs research in lung cancer immunotherapy

Fig. 10.

Top 19 burst keywords in irAE research related to lung cancer immunotherapy (2015–2024). Burst detection analysis identifying keywords with sudden increases in frequency, indicating emerging research topics and novel research directions. Each bar represents a burst keyword, with bar height proportional to the burst strength calculated using Kleinberg’s burst detection algorithm. Keywords are ranked by burst strength, with the strongest bursts indicating the most rapidly emerging research topics. The visualization helps identify cutting-edge research areas and predict future research directions based on burst keyword analysis and temporal patterns

Dominant clusters

Cluster #0 Non-small cell lung cancer (NSCLC) (S = 0.92): This cluster represents a persistent core topic. Keywords: “NSCLC,” “first-line treatment.” Its sustained relevance from the earliest studies to the present underscores the central role of NSCLC (Fig. 11).

Fig. 11.

Timeline view of keyword clusters in irAE research related to lung cancer immunotherapy (2015–2024). Temporal evolution of research topics showing how keyword clusters emerge, develop, and decline over time. Each horizontal band represents a keyword cluster, with the timeline showing the temporal distribution of keyword activity. Colors indicate different research themes, with color intensity representing the relative importance of each theme at different time periods. The timeline view facilitates the identification of research frontiers, emerging trends, and the temporal dynamics of research focus areas

Cluster #1 Interstitial lung disease (S = 0.90): Emerged in 2017 and peaked between 2020 and 2022. Keywords: “Interstitial lung disease,” “pneumonitis.” This trend aligns with increased clinical attention given to immune-mediated pneumonitis. Recent subthemes include diagnostic biomarkers and treatment strategies.

Cluster #2 Neoadjuvant therapy (S = 0.88): Active since 2019. Keywords: “Neoadjuvant immunotherapy,” “resectable NSCLC.” Focus: irAEs in the perioperative setting. However, optimal management remains unclear.

Cluster #3 Solid tumors and Cluster #4 tumor microenvironments (S = 0.87 and 0.86): These clusters provide a mechanistic vocabulary applicable across tumor types. Limited overlap with clinical clusters (e.g., clusters 0, 1, and 2) suggests a need for integration.

Cluster #8 irAEs (S = 0.89): Continuous presence with a secondary burst from 2021 to 2024. Keywords: “immune-related adverse events (irAEs),” “toxicity management.” This confirms irAEs as a standalone research front. This secondary burst reflects a growing emphasis on precision management.

Burst keywords

Keyword burst analysis identified three thematic waves (Fig. 10):

2015–2017: dose-finding and proof-of-concept

Keywords included “metastatic melanoma” (burst = 49.3), “ipilimumab” (burst = 17.4), and “safety” (burst = 18.6). This reflects the early focus of CTLA-4 therapy and initial characterization of immune-related adverse events (irAEs) in patients with melanoma.

2018–2019: phase III registration trials

Keywords included “docetaxel” (burst = 29.5), “PD-1 inhibitors” (burst = 15.2), and “randomized controlled trial” (burst = 9.5). This corresponds to phase III trials comparing PD-1 inhibitors to chemotherapy in non-small cell lung cancer (NSCLC).

2022–2024: combination strategies and precision toxicity

Keywords included “etoposide” (burst = 10.8), which is used in refractory NSCLC combinations, and the re-emergence of “immune-mediated pneumonitis.” This reflects a focus on optimizing combination therapies for hard-to-treat populations and refining toxicity management.

Thematic trajectory

The keyword landscape evolved from “tumor-agnostic mechanisms” (2015–2017) to “NSCLC-specific efficacy validation” (2018–2021) and then to “precision immunotherapy and irAE optimization” (2022–2024). The sustained burst of Cluster #1 and the renewed burst of “irAEs” indicate that toxicity science now coevolves with efficacy research, forming a distinct knowledge growth pole.

Discussion

Evolving knowledge architecture: from efficacy discovery to precision‒toxicity co-optimization

The publication output on irAEs in lung cancer immunotherapy demonstrated a remarkable CAGR of 30.4% from 2015 to 2024. This growth rate exceeds the reported CAGR for immuno-oncology overall (approximately 18%), indicating that irAE research has become an autonomous and rapidly expanding field. The high citation impact (h-index of 246, mean number of citations per paper of 51.35) further supports the significance and novelty of this research domain. Key drivers for this expansion include successive regulatory approvals by the FDA and EMA, which established ICIs as first-line treatments for NSCLC [36, 41, 42], thereby expanding the treated population and the clinical relevance of irAEs. Additional drivers include postmarketing pharmacovigilance requirements generating real-world evidence and the intrinsic biological complexity of irAEs, which provides insights into immune tolerance and tissue-specific regulation [43]. Challenges within the field include knowledge fragmentation across specialized subtopics and a pressing need for standardized definitions and endpoints, as variations in Common Terminology Criteria for Adverse Events (CTCAE) versions can hinder effective data synthesis and cross-study comparisons.

Translating research trends to patient outcomes

The thematic evolution of irAE research, transitioning from initial tumor-agnostic mechanistic studies to focused investigations of organ-directed toxicities, closely parallels advancements in clinical patient management. The sustained research interest in “interstitial lung disease,” for instance, coincided with the development and dissemination of international diagnostic and management guidelines, such as those from oncology societies. The growing emphasis on “irAE subphenotyping” has enabled the development of biomarker-driven interventions, exemplified by the use of IL-6 inhibition for steroid-refractory irAEs [44]. This strategic shift from reactive management to proactive, mechanism-based prevention holds considerable promise for increasing patient survival and quality of life. However, its widespread implementation is currently constrained by the absence of validated point-of-care biomarkers and the need for prospective clinical validation.

Advancing the mechanistic understanding

The cluster representing fundamental immunology exhibited the most rapid citation growth (CAGR of 21% from 2019 to 2024). Foundational research has focused primarily on dysregulated T-cell activation and checkpoint biology. Subsequent investigations explored the roles of tissue-resident memory T (Trm) cells and myeloid inflammation in irAE pathogenesis, with studies linking Trm infiltration to pneumonitis risk. More recent research has elucidated germline genetic risk factors, including specific HLA haplotypes associated with a significantly increased risk of severe irAEs, and the influence of the microbiome on interpatient heterogeneity. These findings signify a progression from descriptive reporting to a deeper interrogation of immunological principles, thereby revealing specific potential targets for therapeutic intervention.

Collaboration networks and the gate-keeping paradox

Despite a trend toward a tripolar global research structure involving the USA, Europe, and Asia–Pacific region, influential gatekeeping roles remain predominantly held by Euro-American institutions. The centrality of INSERM (0.68) was 5.7 times greater than that of the top Chinese institution (0.12), although China contributed 26% of the total publications. This “gate-keeping paradox” stems from several factors: a tendency to publish in regional rather than internationally influential, boundary-spanning journals; a predominance of domestic and monocentric clinical trials among Chinese institutions, in contrast to the highly international collaboration seen in U.S.-led trials; and the underrepresentation of corresponding Chinese authors in the set of publications exhibiting the most rapid citation growth. This pattern suggests that China currently occupies a semi‑peripheral position in the global collaboration network, with limited structural influence despite high output. To transition toward a hub role, Chinese research entities must steward practice-changing datasets, analogous to pivotal trials such as KEYNOTE-024 [26] and CheckMate-227 [27]. This requires leading international trials focused on understudied populations or novel management strategies, which is contingent upon greater regulatory alignment and enhanced data-sharing practices.

Journal ecosystem: specialization and integration

The journal cocitation network exhibited high modularity (Q = 0.7318), facilitating deep expertise within distinct thematic clusters. However, the concomitant low network density (0.0024) indicates a risk of creating insular knowledge silos. A notable symptom of this fragmentation is the limited and delayed integration of mechanistic insights from fundamental science journals into clinical practice guidelines. Boundary-spanning journals, including the Journal of Thoracic Oncology, Annals of Oncology, and the Journal for ImmunoTherapy of Cancer, serve a critical function by bridging these structural holes within the network. Strategic initiatives by these journals, such as curating integrated special issues that pair mechanistic and clinical perspectives, are instrumental in synthesizing knowledge across specialized domains.

Knowledge milestones and thematic evolution

Citation burst analysis revealed three sequential research waves that align closely with key regulatory milestones for ICIs rather than being driven solely by scientific curiosity [20]. The Foundation wave (2013–2017) coincided with the initial approval of PD-1 inhibitors [45]. The validation wave (2018–2021) mapped onto the expansion of ICI indications into first-line NSCLC. The optimization wave (2022–2024) has parallel breakthrough designations for neoadjuvant and perioperative regimens. This policy-responsive pattern likely accelerates clinically relevant progress but may concurrently create research gaps in areas not directly aligned with approval pathways, such as long-term irAEs and rare toxicity subtypes.

The thematic analysis of keyword bursts revealed a conceptual shift. Early studies (2015–2017) were characterized by tumor-agnostic investigations [46, 47]. In contrast, recent research (2022–2024) is increasingly agent-agnostic but intensely focused on organ-specific toxicities [48]. While previous bibliometric studies by Jiang et al. [13] and Gong et al. [14] identified similar foundational research clusters in irAE research, our extended timeframe (2015–2024) captures a crucial recent transition that their analyses did not fully reveal: the emergence of organ-specific toxicity as a distinct research focus and the growing emphasis on precision toxicology. This shift toward subphenotyping irAEs by affected organ systems represents a maturation of the field beyond generalized safety reporting, and it signals the integration of mechanistic insights into clinical management strategies—a trend that our longer observation period uniquely illuminates. This evolution reflects the adoption of advanced organ-focused models and may signal the emergence of a “precision toxicology” paradigm, though its clinical adoption remains speculative and will require prospective validation. The implementation of this paradigm faces barriers, including the high cost of multiomic profiling and the current lack of standardized frameworks for data integration.

Strategic implications for enhancing global research influence

Our analysis reveals a “productivity–influence mismatch” that is not unique to any single country but represents a common challenge for rapidly growing research ecosystems. The case of Chinese irAE research illustrates this phenomenon vividly: despite high publication volume (contributing 26% of global output), its global network influence remains limited, with institutional centrality scores significantly lower than those of established hubs like INSERM (0.12 vs. 0.68). This pattern—high productivity accompanied by peripheral network positioning—offers valuable lessons for any research community seeking to enhance its global standing.

Collaboration network data suggest that transitioning from a peripheral to a hub role requires strategic “precision collaboration”—partnerships with established, high-centrality international institutions that maximize citation impact. For emerging research powerhouses, this means moving beyond bilateral collaborations with core nodes toward leadership in multicenter, intercontinental consortia that generate practice-changing datasets.

Furthermore, emerging research fronts identified by keyword bursts, such as “microbiome” and “biomarker,” represent data-rich niches where regions with large and diverse patient populations hold strategic advantages [49]. Capitalizing on this opportunity requires the establishment of large-scale, longitudinal biobanks that link detailed clinical outcomes with comprehensive multiomic profiling. Generating such practice-changing datasets can reposition any high-output region from a peripheral contributor to a central gatekeeper in the global research landscape, contingent upon sustained investment, standardized data collection protocols, and a commitment to open science.

These observations, while illustrated through the example of Chinese research, are framed in the context of globally applicable strategies for enhancing research influence and collaboration. The principles of precision collaboration, leadership in international trials, and strategic investment in data-rich niches are relevant to researchers and institutions worldwide seeking to increase their global network centrality.

Limitations

This study has several limitations. The exclusive use of the Web of Science Core Collection as a data source may underrepresent non-English publications and gray literature, potentially leading to an overestimation of the influence of English-language research. Consequently, our analysis may overestimate the relative influence of English‑speaking countries and regions (e.g., USA, UK) and underestimate contributions from non‑English scientific production. This bias is inherent to the chosen database and should be kept in mind when comparing publication and citation metrics across countries. This limitation suggests that the observed geographic and institutional dominance of English-speaking regions might be more pronounced in our analysis than in the complete global research effort. Citation burst detection is sensitive to the chosen γ parameter, while robustness checks confirmed the main evolutionary trajectory, minor variations are possible with different parameterizations. The collaboration network analysis is based solely on formal coauthorship data and does not capture informal knowledge exchange through consortia or conferences, potentially omitting influential relationships. Finally, the temporal scope of the analysis (2015–2024) excludes foundational literature published before 2015. This exclusion means that the earliest conceptual and mechanistic pillars of the field are not represented in our mapping of the knowledge structure, which focuses on the modern, high-growth era of ICI application in lung cancer.

Conclusion

The research landscape of irAEs in lung cancer immunotherapy has matured into a dynamic, tripartite ecosystem interlinking clinical practice, mechanistic investigation, and data science. While the United States and institutions such as INSERM maintain central, gatekeeping roles in the global network, our data reveal a notable productivity‑influence mismatch in China’s irAE research output, which is consistent with patterns observed in other rapidly growing research systems. Evolving toward a more polycentric network requires high-output regions to steward practice-changing, intercontinental datasets that address critical unmet needs, such as irAE management in underrepresented populations, and to more effectively bridge mechanistic discoveries with clinical application. Boundary-spanning journals and cross-disciplinary consortia are pivotal in integrating diverse insights to advance toxicity-optimized, precision immunotherapy. Future research should prioritize irAEs in neoadjuvant settings, organ-specific toxicity subphenotyping, and the development of microbiome-driven biomarkers. In these areas, the capacity to generate large, diverse datasets represents a strategic opportunity, particularly for regions such as China, to contribute transformative insights and ultimately help deliver safer and more effective immunotherapy to all lung cancer patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Xiong-Jie Li: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review & editing. Feng-Yue Zhang: Writing-original draft, Writing-review & editing. Xuan Xu: Data curation, Visualization, Writing - review & editing. All the authors read and approved the final manuscript.

Funding

This research did not receive any specific grant from funding.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable. This study used publicly available bibliometric data and did not involve human or animal subjects.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.