Global Research Landscape and Clinical Hotspots in COPD-OSA Overlap Syndrome: A Bibliometric and Visualization Analysis Authors: Shanshan Li, Yanwen Liu, Chenggang Liu Published: 2026 ABSTRACT AbstractBackgroundThis study systematically depicts the global research landscape, knowledge base, and thematic evolution of chronic obstructive pulmonary disease combined with obstructive sleep apnea overlap syndrome (COPD-OSA overlap syndrome), providing a reference for subsequent research and clinical translation.MethodsRelevant literature published between 2000 and 2025 was retrieved from the Web of Science Core Collection, Scopus, and PubMed databases. Records from Web of Science and Scopus were merged and deduplicated to construct the primary bibliometric dataset, while PubMed was analyzed separately as a supplementary source for cross-database comparison using keywords and MeSH terms. Bibliometrix, VOSviewer, and CiteSpace were used to analyze research output, collaboration networks, core sources, co-citation structure, and research hotspots. A latent Dirichlet allocation (LDA) model was applied to identify latent topics and their temporal trends.ResultsThe primary dataset included 1239 publications, with an average annual growth rate of 20.62%; publication volume increased markedly after 2017. The United States, China, and Italy ranked highest in publication count, yet the overall international co-authorship rate was only 6.3%. The field’s knowledge base centers on nocturnal hypoxemia, cardiovascular outcomes, acute exacerbations, and positive airway pressure therapy. The LDA model identified 10 latent topics, with two showing significant upward trends (COVID-19 with OSA; acute cardiovascular events and adipose tissue) and one showing a downward trend (anthropometry and lung function parameters). The 274 PubMed-included articles complemented cross-database comparisons and indicated growing attention to cognitive function, oxygen saturation, and digital screening.ConclusionResearch on COPD-OSA overlap syndrome is shifting from describing disease coexistence toward complex multimorbidity, risk stratification, and individualized management. However, international collaboration and high-level clinical evidence remain insufficient. Future work should standardize case definitions, establish multicenter longitudinal cohorts, and conduct phenotype-based treatment studies.Keywords: chronic obstructive pulmonary disease, obstructive sleep apnea, overlap syndrome, bibliometrics, visualization analysis, latent Dirichlet allocation, PubMed, MeSH Introduction Chronic obstructive pulmonary disease (COPD) and obstructive sleep apnea (OSA) are two common chronic respiratory disorders associated with substantial disease burden. The Global Burden of Disease study indicates that COPD prevalence remains persistently high worldwide,1 and modeling projections suggest that the global number of people with COPD may approach 600 million by 2050.2 Meanwhile, approximately 936 million adults aged 30–69 years are estimated to have mild-to-severe OSA, including about 425 million individuals with moderate-to-severe OSA.3 COPD is characterized by persistent airflow limitation, ventilatory impairment, and gas-exchange abnormalities, whereas OSA is manifested by recurrent upper-airway collapse, intermittent hypoxemia, and sleep fragmentation during sleep. Both diseases are closely associated with cardiovascular disease, metabolic dysfunction, cognitive impairment, reduced quality of life, and increased mortality risk. With the growing burden of population aging, obesity, smoking, and environmental exposures, the clinical implications of the coexistence of these two disorders have attracted increasing attention. The concurrent presence of COPD and OSA is referred to as COPD-OSA overlap syndrome, a concept first proposed by Flenley in 1985.4 Compared with COPD or OSA alone, patients with overlap syndrome typically exhibit more severe and prolonged nocturnal hypoxemia and may also present with hypercapnia, pulmonary hypertension, and higher cardiovascular risk.5 The key clinical question is therefore not simply whether COPD and OSA coexist, but which physiological phenotypes—such as nocturnal hypoxic burden, time with oxygen saturation below 90%, hypercapnia, obesity, and reduced ventilatory reserve—generate additional risk and treatment needs. This dual sustained-intermittent hypoxic insult may plausibly contribute to cardiovascular and metabolic injury through shared pathways such as oxidative stress, systemic inflammation, sympathetic activation, and endothelial dysfunction, although these mechanistic pathways require direct clinical and experimental validation. Therefore, overlap syndrome is not simply a coincidental coexistence of COPD and OSA but a clinically important comorbid phenotype with heterogeneous physiological features, and its management must address both lower-airway obstruction and sleep-related upper-airway collapse. Cohort studies have shown that patients with untreated overlap syndrome have higher risks of all-cause mortality and hospitalization due to acute exacerbation of COPD than patients with COPD alone, and that greater CPAP adherence is associated with improved survival.6,7 However, there remains no consensus regarding the epidemiological characteristics, diagnostic criteria, or optimal therapeutic pathway for overlap syndrome. This uncertainty leads to frequent underdiagnosis or misdiagnosis in clinical practice; most patients with COPD do not undergo routine OSA screening, and vice versa, leaving many patients without targeted positive airway pressure therapy.8,9 Systematic reviews and meta-analyses have also reported substantial heterogeneity in the comorbid prevalence of overlap syndrome across different clinical populations.10 Existing clinical evidence remains largely observational, and reported benefits of CPAP may be influenced by adherence, access to care, baseline disease severity, obesity, smoking, and other health-behavior confounders. In recent years, research in this field has expanded from prevalence surveys and nocturnal hypoxemia characteristics to cardiovascular and metabolic comorbidities, clinical phenotypic heterogeneity, risk prediction, noninvasive ventilation, digital screening, and patient-centered outcomes.11,12 It is increasingly recognized that overlap syndrome is not a homogeneous condition but exhibits pronounced phenotypic heterogeneity. For example, patients may be classified as hypercapnic or normocapnic according to daytime carbon dioxide levels, and these phenotypes differ significantly in clinical characteristics and treatment response. In addition, the effects of overlap syndrome on neurocognitive function have emerged as a growing area of interest. These new directions indicate that the field is moving from broad epidemiological description toward physiological phenotyping and stratified management. Previous bibliometric studies have characterized the research landscape of obstructive sleep apnea and have also been applied to COPD-related comorbid conditions, providing quantitative insights into publication trends, collaboration patterns, and research hotspots.13,14 However, to our knowledge, the global research landscape of COPD-OSA overlap syndrome has not been comprehensively characterized using an integrated multidatabase bibliometric framework. Compared with traditional narrative or systematic reviews, bibliometric analysis enables quantitative mapping of large-scale publication patterns, collaboration networks, knowledge structures, and temporal changes in research topics. It can therefore complement evidence-synthesis approaches by revealing how a research field is organized and how its major and emerging themes evolve over time. Against this background, this study integrated the Web of Science Core Collection, Scopus, and PubMed databases to conduct a bibliometric and visualization analysis of COPD-OSA overlap syndrome research from 2000 to 2025. Bibliometrix, VOSviewer, and CiteSpace were used to systematically evaluate annual output, country and institutional collaboration, core authors and journals, the co-citation-based knowledge base, and keyword evolution. LDA modeling was further applied to identify latent research topics and temporal trends, while PubMed MeSH terms were used for supplementary cross-database consistency assessment. Rather than merely quantifying publication volume, this study aimed to map how evidence, disciplines, institutions, and topics are organized within the COPD-OSA overlap syndrome literature, reveal the evolution of research themes from coexistence-oriented description toward physiological phenotyping, comorbidities, risk outcomes, and personalized management, and provide a basis for future study design and clinical translation. Methodology Data Sources and Search Strategies This study adopted a multidatabase bibliometric framework and systematically retrieved literature on chronic obstructive pulmonary disease (COPD), obstructive sleep apnea (OSA), and COPD-OSA overlap syndrome from the Web of Science Core Collection (WoSCC), Scopus, and PubMed. The search period was set from January 1, 2000, to December 31, 2025, and the final search was conducted on June 1, 2026. To reduce bias arising from differences in database coverage and subject-term indexing, database-specific search strategies were developed according to the field structure of each database Figure 1. Search terms consisted of COPD-related terms and OSA-related terms combined using the Boolean operators “AND” and “OR”. WoSCC and Scopus were searched mainly in the title, abstract, and keyword fields; PubMed was searched using a combination of free-text terms and Medical Subject Headings (MeSH). The complete search strategies for each database are provided in the Supplementary Methods. Workflow of the data search strategy. Records retrieved from WoSCC and Scopus were used to construct the primary bibliometric dataset. The inclusion criteria were as follows: (1) the study involved both COPD and OSA or COPD-OSA overlap syndrome; (2) the publication year was between 2000 and 2025; (3) the document type was Article or Review; (4) the publication was written in English; and (5) complete bibliographic information was available for bibliometric analysis. Conference abstracts, book chapters, errata, retracted publications, and records that mentioned only COPD or OSA without analyzing the relationship between the two conditions were excluded. PubMed was analyzed as a supplementary biomedical data source to compare source journals, core authors, research institutions, keywords, and MeSH term structures with the primary WoSCC-Scopus dataset. PubMed document types were further classified as Article, Review, and Other; Review included reviews, systematic reviews, and meta-analyses, whereas Other mainly included letters, editorials, and comments. All records were exported in full-record and cited-reference formats. Data Integration, Deduplication, and Standardization After importing WoSCC and Scopus records into the R environment, the bibliometrix package was used for format conversion, field matching, and data merging. Duplicate records were first identified using the digital object identifier (DOI). For records without DOI information, title, first author, publication year, and source journal were jointly compared. After automated deduplication, titles and abstracts were further manually checked to remove duplicate entries and records unrelated to the study topic. A primary analytical dataset containing 1239 publications was ultimately generated. PubMed records were analyzed separately, with 274 publications included; these were not merged with the WoSCC-Scopus dataset so that MeSH indexing information could be retained and consistency across databases could be evaluated. Before network analysis, author, institution, country, journal, and keyword fields were standardized. Author names were unified as surname plus initials; full institutional names, abbreviations, and affiliated subunits referring to the same institution or campus were merged when the affiliation field unambiguously supported merging. Parent university systems and specific campuses were not automatically collapsed when campus attribution could not be reliably determined from the exported affiliation string. Country names were corrected using standardized English nomenclature. Keywords were normalized irrespective of case, and singular/plural forms, hyphenation, abbreviations, and full expressions were harmonized. Synonymous terms such as COPD and chronic obstructive pulmonary disease, as well as OSA and obstructive sleep apnea, were merged according to the purpose of each analysis. Generic terms with limited discriminatory value, abnormal characters generated during database conversion, and semantically meaningless terms were removed after manual review. To ensure reproducibility, the raw dataset, deduplicated dataset, and standardized dataset were saved separately after data cleaning. The merging rules for keywords and institutional names were recorded in a thesaurus and applied consistently across all network analyses. Analysis Tools and Visualization Methods Descriptive bibliometric analyses were conducted in R software (version 4.5.1) using the bibliometrix package. The analyzed indicators included annual publication output, annual growth rate, total citations, average citations per document, author and institutional productivity, country scientific production, source journals, and highly cited papers. Author, journal, and document impact were evaluated using publication counts, total citations, and h-index metrics. Country collaboration was categorized into single-country publications (SCP) and multiple-country publications (MCP), and the MCP proportion was used to quantify the degree of international collaboration. Full counting was used to describe overall output, whereas fractional counting was used to reduce the influence of multi-author or multi-institutional collaboration on rankings. The ratio of local citations to global citations (LC/GC) was used to compare the relative influence of publications within the COPD-OSA research field and in the broader scholarly network. VOSviewer (version 1.6.20) was used to construct author collaboration, institutional collaboration, keyword co-occurrence, author co-citation, reference co-citation, and bibliographic coupling networks. In these networks, node size represented publication count, occurrence frequency, or citation frequency; edges indicated collaboration, co-occurrence, co-citation, or coupling relationships; edge thickness represented total link strength; and colors indicated clustering results. Only nodes meeting the prespecified minimum occurrence thresholds were included, with thresholds set according to the analytical object and reported in the corresponding figure legends. Keyword temporal overlay maps were colored according to average publication year to identify early foundational topics and emerging research hotspots. CiteSpace (version 6.4.R1) was used to analyze reference co-citation networks, keyword clusters, timeline distributions, burst keywords, and burst references. The time slice was set to one year, and node types were selected as Reference, Keyword, Author, Institution, or Country according to the analysis purpose. Network pruning was performed using Pathfinder and pruning sliced networks. Burst detection was used to identify keywords and references that showed rapid increases in frequency or citation intensity during specific periods, and burst strength and duration were used to evaluate the activity of research frontiers. A dual-map overlay of journals was used to visualize disciplinary citation trajectories between citing and cited journals, with curve width representing the relative strength of citation links. To further identify latent semantic structures not readily captured by conventional co-occurrence analysis, latent Dirichlet allocation (LDA) was applied to titles, abstracts, and keywords. Text preprocessing included lowercasing, removal of punctuation and stop words, lemmatization, and extraction of unigrams and bigrams. The final analysis used a 10-topic LDA model. Model-selection diagnostics, including coherence, perplexity, and stability checks, were reported in the Supplementary Methods and were kept consistent with this 10-topic solution. PubMed data were analyzed using the same descriptive workflow, and co-occurrence networks were constructed based on author keywords and MeSH terms. Node size represented term frequency, edges represented co-occurrence relationships, and overlay colors represented average publication year. Results Publication Output and General Characteristics After retrieval from Web of Science and Scopus and subsequent deduplication, 1239 publications published between 2000 and 2025 were included, with an annual growth rate of 20.62%. These comprised 979 Articles and 260 Reviews (Table 1). Annual publication output showed an overall sustained upward trend (Figure 2A) and could be divided into two stages: a relatively slow growth phase from 2000 to 2016, during which the annual number of publications remained below 50, and a markedly accelerated growth phase after 2017, reaching 217 publications in 2025. The citation indicators in Figure 2B should be interpreted separately: mean total citations per article were higher for early publication years, with 275 in 2001 and 136.56 and 159.95 in 2009 and 2010, respectively, whereas mean annual citations per article reached 7.59 and 9.41 in 2009 and 2010, respectively. These early-year citation advantages may partly reflect longer citation windows. Table 1Main InformationDescriptionResultsMain Information About DataTimespan2000:2025Sources (Journals)509Documents1239Annual Growth Rate %20.62Document Average Age6.03Average citations per doc27.63References10048Document ContentsKeywords Plus (ID)7410Author’s Keywords (DE)2329AuthorsAuthors7225Authors of single-authored docs38Authors CollaborationSingle-authored docs43Co-Authors per Doc7.52International co-authorships %6.295Document TypesArticle979Review260 Main Information (A) Annual publication output; (B) citation indicators over time, including mean total citations per article and mean annual citations per article; (C) publications by corresponding author country; (D) geographic distribution of corresponding author countries. As shown in Figure 2C and D, the United States had the highest publication output, followed by China and Italy. The analysis of corresponding author countries (Table 2) showed that the MCP proportion in the United States was 4.3%, which was lower than the overall level. Greece (17.9%), Canada (15.9%), Australia (13.8%), and the United Kingdom (13.2%) showed relatively higher MCP proportions. Table 2Top Countries by Corresponding Author PublicationsCountryArticlesSCPMCPMCP %USA369353164.3China12411775.6Italy817833.7Spain635946.3Canada4437715.9Germany393725.1France3834410.5United Kingdom3833513.2Australia2925413.8Greece2823517.9 Top Countries by Corresponding Author Publications Contributions by Countries, Regions, and Institutions The international collaboration network (Figure 3A) showed that the strongest collaboration was between the United States and China, followed by the United States and Canada, the United States and Spain, Italy and Greece, the United States and Italy, and France and Switzerland. The overall proportion of international co-authorship was 6.3% (Figure 3C). The United States and China had international collaboration proportions of 4.3% and 5.6%, respectively, whereas Greece and Canada exhibited relatively higher proportions of international collaboration. (A) International collaboration network; (B) funding support distribution; (C) chord diagram of international collaboration. Regarding institutional output (Table 3), Democritus University of Thrace ranked highest in publications, with the University of California San Diego, Harvard Medical School, and Mayo Clinic ranking second to fourth. Multiple University of California-related records appeared among the top ten. In accordance with the institutional standardization rules, University of California San Diego was retained as a specific campus, whereas the generic University of California entry was retained separately because those records could not be reliably assigned to a single campus from the exported affiliation fields. Among Chinese institutions, Capital Medical University and Soochow University entered the list of high-output institutions, reflecting that China has established a certain research presence in this area. In addition, the University of Crete (Greece), McMaster University (Canada), and University College Dublin (Ireland) also ranked within the top ten. Table 3Top 10 Most Productive InstitutionsCountryAffiliationArticlesGreeceDemocritus University of Thrace56United StatesUniversity of California San Diego45United StatesHarvard Medical School39United StatesMayo Clinic35United StatesUniversity of California33ChinaCapital Medical University31GreeceUniversity of Crete24CanadaMcmaster University20ChinaSoochow University17IrelandUniversity College Dublin17 Top 10 Most Productive Institutions The funding-source distribution analysis shown in Figure 3B indicated that research on COPD–OSA overlap syndrome received support from multiple sources, including corporations, national research programs, and medical research foundations. Core Authors and Collaboration Networks The ranking of core authors by publication output (Table 4) showed that Zhang Y had the highest number of publications (47), followed by Malhotra A and McNicholas W. Both Malhotra A and McNicholas W had total citations of 1618 and h-index values of 16. Highly productive authors were distributed across China, the United States, France, Greece, and Ireland, with major affiliations including Peking University First Hospital, University of California San Diego, University College Dublin, University Grenoble Alpes, and Democritus University of Thrace. Table 4Top 10 Most Productive AuthorsAuthorArticlesCountryH_indexTCNPPY_startAffilliationsZhang Y47China13666472009Peking University First HospitalMalhotra A27USA161618272010University of California San DiegoMcnicholas W23Ireland161618232009University College DublinPepin J17France13953172013University Grenoble AlpesSchiza S16Greece9437162007University of CreteSteiropoulos P16Greece9319162016Democritus University of ThraceOwens R15USA11621152010University of California San DiegoWang Y15 China9293152015Chung Shan Medical University HospitalBailly S14France10393142020Laboratoire Hypoxie et Physiopathologies Cardiovasculaires et RespiratoiresTamisier R13France11398132014Laboratoire Hypoxie et Physiopathologies Cardiovasculaires et Respiratoires Top 10 Most Productive Authors Figure 4A showed that several relatively stable collaboration clusters had formed in this field, including a French cluster centered on Pepin J, Tamisier R, and Bailly S; a Greek cluster centered on Steiropoulos P and Schiza S; and a Chinese cluster represented by Zhang Y, Wang H, and Wang Y. More cross-national links were observed between the French and Greek clusters, whereas the Chinese author cluster had relatively limited connections with clusters from other countries. Figure 4B showed that Marin JM, McNicholas WT, Chaouat A, and Flenley DC were located in the core area of the network, while Young T, Malhotra A, and Ryan S also served as important nodes. Highly co-cited authors were mainly from Europe and North America. (A) Author collaboration network; (B) author co-citation network. Core Journals, Highly Cited Papers, and Knowledge Base Table 5 lists the top 10 source journals by publication output and their key impact indicators. Sleep and Breathing published the largest number of papers, followed by Journal of Clinical Sleep Medicine and Journal of Clinical Medicine. In terms of academic influence, CHEST published only 16 relevant papers but accumulated 1452 citations, with approximately 90 citations per article, an h-index of 14, a JCR Q1 ranking, and an impact factor of 9.8, highlighting its authoritative leadership in this field. Journal of Clinical Sleep Medicine had 1001 total citations and an h-index of 18, also indicating strong scholarly influence. Sleep Medicine and Respiratory Medicine were also important source journals in this field. Regarding publishers, Springer, Elsevier, the American Academy of Sleep Medicine, Taylor and Francis Group, and other publishers/societies were represented among the top 10. Figure 5A showed that Sleep and Breathing and Journal of Clinical Sleep Medicine began to increase rapidly around 2015. Table 5Top 10 Most Productive JournalsPublisherSourcesArticlesH_indexTCJournal Citation Reports QuartileImpact FactorPY_startSpringerSleep and Breathing4717830Q422002American Academy of Sleep MedicineJournal of Clinical Sleep Medicine37181001Q32.92008MdpiJournal of Clinical Medicine349376Q33.92019ElsevierSleep Medicine3215614Q24.12008ElsevierRespiratory Medicine2114693Q23.12007SpringerBmc Pulmonary Medicine208236Q32.82014Taylor & Francis GroupInternational Journal of Chronic Obstructive Pulmonary Disease179266Q33.12006Public Library of SciencePlos One1711349Q32.62015American College of Chest PhysiciansChest16141452Q19.82001ElsevierSleep Medicine Clinics16578Q32.92008 Top 10 Most Productive Journals (A) Cumulative production of core journals over time; (B) references with the strongest citation bursts. Red bars indicate the periods of citation bursts, whereas blue bars represent the non-burst periods within the overall study timespan. Table 6 analysis shows that Bednarek M (2005, Respiration) has the highest local citation count (36), with an LC/GC ratio of 27.69%; Mermigkis C (2007, International Journal of Clinical Practice) has a local citation count of 29, LC/GC ratio 48.33%; Adler D (2020, PLOS ONE) has a local citation count of 24, LC/GC ratio 52.17%. Agusti A (2011, European Respiratory Review) has a local citation count of 15 and a global citation count of 215, with an LC/GC ratio of 6.98%. Table 6Most Locally Cited PapersPaperDOILocal CitationsGlobal CitationsLC/GC Ratio (%)Normalized Local CitationsNormalized Global CitationsBednarek M, 2005, Respiration10.1159/00008404436.00130.0027.694.801.47Mermigkis C, 2007, Int J Clin Pract10.1111/j.1742–1241.2006.01213.x29.0060.0048.3311.231.09Adler D, 2020, Plos One10.1371/journal.pone.023533124.0046.0052.1729.541.65Budhiraja R, 2015, J Clin Sleep Med10.5664/jcsm.454023.00175.0013.1418.403.98Greenberg-Dotan S, 2014, Sleep Breath10.1007/s11325-013-0850-320.0039.0051.2812.960.71Lacedonia D, 2018, Clin Respir J10.1111/crj.1275417.0051.0033.3319.521.66Zhang P, 2022, Bmc Pulm Med10.1186/s12890-021-01780-416.0031.0051.6132.701.99Agusti A, 2011, Eur Respir Rev10.1183/09059180.0000431115.00215.006.9812.002.93Naranjo M, 2020, Chest10.1016/j.chest.2020.03.03613.0057.0022.8116.002.04Jaoude P, 2014, Lung10.1007/s00408-014-9555-z12.0040.0030.007.780.73 Most Locally Cited Papers The burst analysis in Figure 5B showed that Marin JM (2010, American Journal of Respiratory and Critical Care Medicine) had the highest burst strength, with a burst period from 2016 to 2024. Shawon MSR (2017, Sleep Medicine Reviews) and McNicholas WT (2017, CHEST) had burst strengths of 24.05 and 22.93, respectively, with burst periods from 2018 to 2025 and from 2018 to 2024. Flenley DC (1985, Clinics in Chest Medicine) had a burst strength of 15.52 during 2016–2025. The burst periods of Adler D (2020, PLOS ONE) and Benjafield AV (2019, The Lancet Respiratory Medicine) extended to 2025. Figure 6A showed that Marin JM (2010), Flenley DC (1985), and Chaouat A (1995) were located in the core area and were closely connected with nodes such as Sanders MH (2003) and McNicholas WT (2009). The network mainly formed three clusters related to epidemiology and cardiovascular outcomes, clinical diagnosis and treatment, and the pathophysiology of sleep apnea. In Figure 6B, recent publications by authors such as McNicholas W, Owens R, and Sunwoo B constituted multiple interconnected research clusters. Figure 6C illustrated the knowledge flow between citing and cited journals. The main citation trajectories extended from journals in medicine, clinical medicine, molecular biology, and immunology to clusters in health, nursing, medicine, molecular biology, and genetics. Among them, the pathways from Medicine/Medical/Clinical to Health/Nursing/Medicine and Molecular/Biology/Genetics were the most prominent, indicating that publications in this field are mainly published in clinical medicine journals and that their knowledge base draws on both clinical health sciences and molecular biology research. (A) Citation analysis network; (B) bibliographic coupling network; (C) dual-map overlay of journals. Keyword Co-Occurrence, Clustering, and Evolutionary Analysis Figure 7A showed that prevalence, COPD, obstructive sleep apnea, pulmonary disease, outcomes, positive airway pressure, mortality, risk, and overlap syndrome were the main high-frequency terms in this field. The thematic trend analysis (Figure 7B) indicated that early research primarily focused on oxygen partial pressure, carbon dioxide partial pressure, pulmonary hemodynamics, and sleep apnea syndrome; subsequently, the focus gradually shifted toward hypoventilation, systemic inflammation, pathogenesis, and sleep-disordered breathing. In recent years, comorbidities, obesity, etiology, pharmacotherapy, and cohort and retrospective studies have increased markedly. Figure 7C further illustrated stage-specific changes in research priorities: from 2000 to 2005, research mainly focused on hypoxemia, hypoventilation, pulmonary hypertension, and mortality; from 2005 to 2010, it expanded to obesity, cardiovascular comorbidities, and insomnia; from 2010 to 2015, the focus shifted to prevalence, epidemiology, CPAP, and noninvasive ventilation; after 2015, treatment adherence, clinical outcomes, phenotypes, quality of life, readmission, and machine learning gradually attracted attention. Figure 7D showed that obstructive sleep apnea, chronic obstructive pulmonary disease, overlap syndrome, positive airway pressure, prevalence, mortality, and risk constituted the major hub nodes in the network. Figure 7E formed six interconnected clusters. The major themes included COPD-related respiratory pathophysiology and pulmonary function; cardiovascular and metabolic comorbidities; epidemiology, risk factors, and clinical study design; obesity and clinical outcomes; prevalence, mortality risk, and PAP therapy in overlap syndrome; and polysomnography, apnea-hypopnea index, and sleep symptom assessment. Burst analysis (Figure 7F) showed that hemodynamics had the longest burst duration (2001–2010), whereas disorders had the highest burst strength (3.65), followed by continuous positive airway pressure (3.34), daytime sleepiness (3.25), noninvasive ventilation (3.07), pathophysiology (2.64), and C-reactive protein (2.51). Among these, risk factors (2022–2025) and health (2023–2025) continued to burst until the end of the study period, representing recently active research directions. (A) Word cloud of high-frequency keywords; (B) thematic evolution trend of keywords; (C) keyword timeline map; (D) temporal overlay network of keyword co-occurrence; (E) keyword co-occurrence clustering network (minimum occurrence threshold in VOSviewer = 10); (F) keywords with the strongest citation bursts. Red bars indicate the periods of citation bursts. Latent Structure and Temporal Evolution Revealed by LDA Topic Modeling To further identify latent semantic structures that traditional keyword co-occurrence analyses may miss, this study constructed an LDA topic model based on titles, abstracts, and keywords, ultimately identifying 10 latent research topics. The overall prevalence, number of documents, and high-weight terms for each LDA topic are provided in Supplementary Table 1. The average proportion of each topic across the entire corpus is relatively similar, indicating that research on COPD-OSA overlap syndrome is not dominated by a single theme but is composed of multiple parallel research directions. Based on high-weight terms and representative literature, these topics broadly cover basic respiratory physiology and animal experiments; anthropometric characteristics and pulmonary function in humans; asthma and complex comorbidities; noninvasive ventilation and cardiovascular events; sleep monitoring and psychiatric comorbidities; perioperative management and cardiopulmonary comorbidities; and COVID-19 with concurrent OSA, among others. Figure 8A shows that the 10 topics exhibit certain semantic distribution differences in the two-dimensional space formed by PC1 and PC2. PC1 and PC2 explain 11.9% and 11.7% of the variance, respectively, with a cumulative explained variance of 23.6%. Therefore, this plot is mainly intended to assist in observing relative semantic proximity among topics and the two-dimensional distances should not be overinterpreted as strict topic boundaries. Overall, Topic 1 and Topic 9 lie in the positive region of PC1, suggesting some association with certain clinical phenotypes and pulmonary function-related content; Topic 3 and Topic 10 lie in the negative region of PC1, possibly reflecting another group of comorbidity- or clinical problem-related themes; Topic 6 is prominent in the negative region of PC2, whereas Topics 2, 5, 7, and 8 are relatively close to the positive region of PC2, indicating a degree of clustering among these topics in the latent semantic structure. The slope, P value, R-squared value, and trend classification for each topic are summarized in Supplementary Table 2. Figure 8C reveals temporal dynamics of the different topics from 2000 to 2025. Analysis in Figure 8B shows a significant upward trend for Topic 7, indicating a marked recent increase in research on COVID-19 with concurrent OSA; Topic 10 also shows a significant upward trend, indicating growing attention to acute cardiovascular events, adipose tissue, and related metabolic risk topics. Conversely, Topic 9 shows a significant downward trend, indicating a gradual decline in the relative share of traditional descriptive studies centered on anthropometric characteristics and pulmonary function parameters within the overall knowledge structure. The remaining seven topics did not reach statistically significant change, suggesting that directions such as basic respiratory physiology, sleep monitoring, complex comorbidities, perioperative management, and ventilatory therapies remained relatively stable during the study period. (A) Principal component analysis biplot of the 10 topics, with colors representing four clusters (PC1: 11.9%; PC2: 11.7%). (B) Slopes and significance levels of temporal trends for each topic (blue: significant increase; red: significant decrease; gray: no significant change; dashed line: P = 0.05 threshold). (C) Temporal trends in the average proportion of each topic. Overall, the LDA results indicate that research on COPD-OSA overlap syndrome is shifting from an early research framework dominated by descriptions of disease coexistence, anthropometry, and pulmonary function assessment toward emerging directions such as COVID-19–related impacts, cardio-metabolic risk, adipose tissue–related mechanisms, and complex clinical phenotypes. However, LDA topics reflect latent semantic structures and relative changes in topic proportions within the literature text; they do not directly demonstrate disease mechanisms or causal treatment relationships. Therefore, interpretations of rising topics should be confined to research attention and the evolution of the knowledge structure; related mechanisms and clinical significance still require validation through prospective cohort studies and interventional research. Bibliometric Features and Thematic Structure of PubMed Data The PubMed dataset included 274 publications published between 2000 and 2025, involving 98 source journals, including 180 Articles, 81 Reviews, and 13 other document types (Table 7). Figure 9A showed that Sleep and Breathing had the highest publication output, followed by Sleep Medicine and International Journal of Chronic Obstructive Pulmonary Disease. Figure 9B indicated that Malhotra A had the highest publication output, followed by McNicholas WT, Owens RL, and Zhang J. Figure 9C showed that Democritus University of Thrace ranked first in the PubMed institutional output analysis, followed by the University of California and Tianjin Medical University General Hospital. Capital Medical University, Renmin Hospital of Wuhan University, and King’s College London were also among the productive institutions in the PubMed dataset. Figure 9D formed three interconnected thematic groups, mainly involving comorbidities and risk factors, respiratory pathophysiology and ventilatory support, and clinical outcomes such as hospitalization and mortality. Figure 9E showed that Pulmonary Disease, Chronic Obstructive and Sleep Apnea, Obstructive were core nodes and were closely connected with MeSH terms such as Comorbidity, Continuous Positive Airway Pressure, Polysomnography, Hypoxia, Hypercapnia, and Noninvasive Ventilation. Overlay visualization showed that Biomarkers, Cognitive Dysfunction, Oxygen Saturation, and Autoimmune Diseases had relatively later average publication years. Table 7Main Characteristics of the PubMed DatasetDescriptionResultsTimespan2000:2025Sources (Journals)98Single-authored docs13Document ContentsIndexed terms911International co-authorships %13.14Documents274Document TypesArticle180Review81Other13Document Average Age7.79Co-Authors per Doc5.76Authors of single-authored docs10Authors CollaborationAuthors1261Author’s Keywords (DE)911Annual Growth Rate %12.95 Main Characteristics of the PubMed Dataset (A) Top 10 most productive journals in PubMed; (B) top 10 most productive authors in PubMed; (C) top 10 most productive institutions in PubMed; (D) keyword co-occurrence network in PubMed; (E) MeSH term co-occurrence network in PubMed. The PubMed analyses were broadly consistent with the primary dataset in terms of core journals, productive authors/institutions, and keyword structures, supporting the robustness of the main thematic findings. Meanwhile, the later average publication years of Cognitive Dysfunction and Oxygen Saturation in the MeSH network suggest that these directions have received additional attention in recent years. Discussion This study found that the United States, China, and several European countries constitute the major research forces in this field, although the overall level of international collaboration remains limited. Although the United States and China had high publication output, their proportions of internationally collaborative papers were relatively low. In contrast, Greece, Canada, the United Kingdom, and Australia exhibited more active cross-national collaboration. Author and institutional networks also formed several geographically oriented clusters, and Chinese research teams had relatively limited connections with the European and North American core networks. The geographic distribution of research also indicates relatively limited representation from many developing countries. This imbalance deserves attention because COPD-OSA overlap syndrome remains underrecognized in clinical practice, partly owing to the substantial underdiagnosis of both COPD and OSA.15 This challenge may be particularly relevant in resource-limited settings, where access to sleep diagnostic services, polysomnography, and integrated respiratory-sleep care may be restricted. Consequently, the current evidence base may not fully reflect the disease burden, clinical phenotypes, and management challenges in these populations. Future studies should therefore strengthen research capacity and multicenter collaboration in underrepresented regions to improve the global representativeness of evidence on COPD-OSA overlap syndrome. This collaboration pattern has important methodological implications. Previous systematic reviews have found substantial between-study variation in the prevalence and clinical outcomes of overlap syndrome, and this heterogeneity is related to case sources, OSA diagnostic thresholds, COPD definitions, obesity composition, and sleep-monitoring methods.16 Therefore, single-center or single-region studies cannot easily determine whether observed differences originate from disease biology or from case selection and diagnostic workflows. Future studies should establish prospective cohorts covering diverse regions and ethnic groups and harmonize definitions of pulmonary function, sleep-respiratory events, nocturnal hypoxic burden, and clinical outcomes. The core value of strengthening international collaboration lies not merely in increasing publication counts but in enhancing the representativeness and external validity of study findings. The bibliometric findings showed that overlap syndrome, as an independent keyword, appeared throughout the entire study period and served as a hub connecting the COPD and OSA clusters in the co-occurrence network. This suggests that overlap itself has evolved into a research concept with independent scholarly value. Its clinical importance arises from the interactive amplification of hypoxia between the two diseases: daytime hypoxemia caused by COPD is superimposed on nocturnal intermittent hypoxemia caused by OSA, producing more severe nocturnal oxygen desaturation than either disease alone. This synergistic effect substantially amplifies cardiovascular risk, and patients with overlap syndrome have significantly higher risks of pulmonary hypertension, heart failure, and arrhythmia than patients with either disease alone.17 Therefore, treating overlap syndrome as an integrated research unit is important for understanding its clinical heterogeneity, potential pathophysiological interactions, and phenotype-specific management needs. Co-citation and burst analyses showed that mortality, acute exacerbation of COPD, and positive airway pressure therapy in overlap syndrome have long occupied central positions in the knowledge network. The landmark cohort study by Marin et al remains one of the most important co-cited nodes in this field.6 However, recent research has gradually shifted from confirming whether CPAP is beneficial to identifying which patients may derive greater benefit. In a United States Medicare population, Singh et al found that after CPAP initiation, the proportion of hospitalizations due to COPD-related diseases decreased compared with the pre-treatment period, whereas all-cause hospitalizations and emergency department visits did not improve to the same extent.18 Jaoude and El-Solh further reported that poorer pulmonary function, persistent smoking, and lower CPAP adherence were associated with acute exacerbations of COPD, whereas the degree of airflow limitation was not significantly associated with the apnea-hypopnea index.19 These findings are consistent with the sustained activity of treatment response, acute exacerbation, and risk factor themes observed in the present study, and they also indicate that using the severity of OSA or COPD alone may be insufficient to predict outcomes. Phenotypic differences may be an important explanation for inconsistent treatment effects. Previous cohort studies have shown that the survival benefit of CPAP is more pronounced in hypercapnic patients,20 while other research has found that the effects of CPAP on arterial blood gas and pulmonary function may be influenced by baseline airflow limitation.21 Therefore, future studies should move beyond simple comparisons of CPAP use versus nonuse; stratification should be performed according to hypercapnia, hypoxic burden, obesity hypoventilation, airflow limitation, and symptom burden, and the long-term effects of CPAP, bilevel positive airway pressure, and other noninvasive ventilation strategies should be compared. Keyword and MeSH networks showed that cardiovascular disease, hypertension, heart failure, hospitalization, and mortality were closely linked with COPD, OSA, and overlap syndrome. This indicates that cardiovascular outcomes have become a core theme connecting respiratory pathophysiology with long-term prognosis. Existing clinical studies suggest that nocturnal hypoxic burden may be more directly aligned with cardiovascular risk pathways than the apnea-hypopnea index alone. In a retrospective cohort, Akinnusi et al found that the time spent with nocturnal oxygen saturation below 90% was an independent predictor of incident atrial fibrillation, whereas better CPAP adherence was associated with lower atrial fibrillation risk; patients who developed atrial fibrillation also had higher risks of hospitalization and mortality.22 In addition, a review of sleep disorders and acute exacerbations of COPD showed that both OSA and insomnia may increase susceptibility to exacerbations, suggesting that sleep quality, hypoxemia, and respiratory events may influence COPD outcomes through different pathways.23 Therefore, future risk stratification should not rely solely on the presence or absence of overlap syndrome or on the apnea-hypopnea index. Nocturnal hypoxic burden, cumulative time below 90% oxygen saturation, hypercapnia, heart rate variability, and pulmonary artery pressure may provide more clinically informative data. Longitudinally linking these indicators with cardiovascular events, acute exacerbations, and mortality may facilitate the development of a risk-assessment framework that shifts from disease diagnosis to physiological phenotyping. The LDA analysis showed that the relative proportion of traditional epidemiological topics centered on anthropometric features and pulmonary function parameters decreased, whereas topics related to COVID-19 with OSA and acute cardiovascular events and adipose tissue showed significant upward trends. This change suggests that the research focus is expanding from descriptive epidemiology toward emerging comorbidities and complex clinical outcomes. Macrea et al found that among patients with COPD and severe comorbidities, those with overlap syndrome may have an elevated peripheral neutrophil proportion; however, the sample size was small, and it remains unclear whether this represents an independent inflammatory biomarker.24 Recent bioinformatics research has further screened common differentially expressed genes from COPD and OSA gene-expression data and attempted to validate candidate hub genes in independent datasets.25 These studies are consistent with a broader emerging interest in biomarker-oriented and molecular research. However, the significant LDA slopes in the present analysis directly support upward trends only for COVID-19 with OSA and acute cardiovascular/adipose tissue-related topics. Therefore, biomarker-related findings should be interpreted as an emerging research direction suggested by recent studies and term-level signals, rather than as an established mechanistic theme confirmed by LDA alone. Notably, obesity, one of the major drivers of COPD–OSA overlap syndrome, has recently emerged as an important therapeutic target. The Phase 3 SURMOUNT-OSA trial demonstrated that tirzepatide significantly reduced the apnea–hypopnea index (AHI), body weight, hypoxic burden, high-sensitivity C-reactive protein levels, and systolic blood pressure after 52 weeks in adults with obesity and moderate-to-severe OSA, while also improving patient-reported sleep outcomes.26 Although these findings were obtained in patients with OSA rather than COPD–OSA overlap syndrome, they provide an important proof of concept that obesity-targeted pharmacotherapy may represent a promising adjunctive strategy for overlap syndrome, particularly in patients with an obesity-predominant phenotype. However, whether these benefits can be translated into improvements in COPD-related outcomes, exacerbation risk, cardiopulmonary function, or long-term prognosis remains unknown. Future prospective studies are therefore warranted to evaluate the efficacy and safety of tirzepatide and other anti-obesity therapies in COPD–OSA overlap syndrome and to determine their potential additive effects when combined with positive airway pressure therapy. In recent years, researchers have attempted to distinguish an OSA-dominant hypoxic phenotype (high nocturnal hypoxic burden with relatively preserved daytime ventilation) from a COPD-dominant ventilatory phenotype (daytime hypoxemia with hypercapnia) in overlap syndrome. These two phenotypes may differ fundamentally in their responses to PAP, oxygen therapy, and pulmonary rehabilitation. In addition, glucagon-like peptide-1 (GLP-1) receptor agonists have shown potential for improving the apnea-hypopnea index in patients with obesity and OSA, providing a new pharmacological direction for treating the obesity phenotype in overlap syndrome; however, the effects of these agents on COPD-related outcomes remain unclear. Notably, women with OSA often present with atypical symptoms, such as insomnia and fatigue rather than sleepiness, which may affect screening rates and diagnostic accuracy for overlap syndrome among female patients with COPD. Systematic studies targeting these special populations remain relatively scarce in the current literature. Bibliographic coupling and recent subject terms indicated that research outcomes are expanding from mortality, hospitalization, and acute exacerbation to quality of life, cognitive function, psychological and psychiatric status, exercise capacity, and fall risk. This transition suggests that evaluation is gradually shifting from disease control to overall patient function. In a cross-sectional study, De Censo et al found that patients with COPD and comorbid moderate-to-severe OSA had more pronounced postural balance impairment and reported more falls during follow-up.12 This finding supports the emergence of quality-of-life, functional-status, and nonrespiratory outcome themes in the present study. However, the evidence came from a relatively small sample, and the influences of age, medication, muscle function, and comorbidities could not be fully excluded. Therefore, it cannot yet be concluded that OSA directly causes functional deterioration in patients with COPD. Future research should simultaneously evaluate sleep architecture, hypoxic burden, exercise tolerance, cognitive function, falls, quality of life, and treatment adherence, and should determine whether these outcomes can be modified by CPAP, noninvasive ventilation, oxygen therapy, or pulmonary rehabilitation. Patient-reported outcomes should be incorporated into core outcome sets together with pulmonary function and sleep-monitoring indicators to enable a more comprehensive evaluation of treatment benefit. This study identified emerging themes such as machine learning, wearable monitoring, and risk prediction, indicating that digital technologies are beginning to enter overlap syndrome research. Zhang et al used an audio-enabled smartwatch to simultaneously monitor blood oxygen, heart rate, respiratory rate, and cough and forced-expiration sounds, preliminarily exploring its feasibility for OSA screening in patients with COPD.27 In addition, multicenter prospective study protocols have proposed the use of the medical Internet of Things to construct tiered screening and early-warning systems for COPD with comorbid OSA.28 These studies provide new technical pathways for expanding OSA screening coverage; however, current evidence remains largely exploratory or at the proof-of-concept stage. Model performance may be influenced by sample size, devices, ambient noise, and patient characteristics. Future studies should conduct external validation and compare incremental value against the STOP-Bang questionnaire, Epworth Sleepiness Scale, pulmonary function, and nocturnal oximetry. The intended use of digital tools should also be clearly defined, such as preliminary screening, referral, or follow-up, rather than replacing polysomnography and standardized clinical diagnosis. Taken together, the dual-map overlay of journals showed that the main knowledge flow in this field occurs among clinical medicine, health sciences, and molecular biology, whereas knowledge input from psychology, rehabilitation, and social sciences remains relatively limited. This indicates that overlap syndrome research has acquired interdisciplinary features, but basic mechanisms, clinical treatment, and long-term management have not yet been sufficiently integrated. Based on the bibliometric findings of this study and supplementary clinical evidence, four directions warrant continued validation: first, establishing a unified case definition and core outcome set; second, building multicenter longitudinal cohorts covering diverse populations; third, conducting treatment stratification based on hypoxemia, hypercapnia, obesity, and exacerbation propensity; and fourth, evaluating whether biomarkers and digital screening can improve conventional risk models. Recent research has also proposed OSA as a treatable trait in COPD and explored the potential of combining CPAP with inhalation therapy, but current evidence remains insufficient to support routine treatment recommendations.29 Therefore, the above directions should be interpreted as hypotheses generated jointly from bibliometric signals and existing clinical studies, rather than as clinical conclusions directly demonstrated by the present study. The absence of a marked increase in large randomized controlled trial-related themes further suggests that high-level clinical evidence remains limited in this field. Limitations This study has several limitations. First, the primary data were obtained from the Web of Science Core Collection and Scopus and were restricted to English-language publications, which may have led to the omission of non-English studies, regional journals, and gray literature. Web of Science and Scopus differ substantially in journal coverage and regional representation,30 whereas PubMed focuses more on biomedical literature and MeSH indexing. Therefore, consistency across databases cannot completely eliminate database-selection bias. Second, publication counts, citation counts, and h-index values mainly measure scholarly dissemination and are not equivalent to study quality or level of clinical evidence. Citation indicators may be affected by publication year, journal impact, topic popularity, and author and journal self-citations. Because self-citations were not fully excluded, the influence of some authors, institutions, or journals may have been overestimated. Third, citation analysis is subject to time-accumulation effects. Earlier publications have longer citation windows, whereas recent studies may be underestimated because citation accumulation is still insufficient. Burst analysis and average publication year help identify emerging topics but cannot determine whether these topics will persist over the long term. Fourth, author, institution, and keyword fields contain abbreviations, spelling variants, and name variants. Although standardization and synonym merging were performed, author homonyms, institutional splitting, or inappropriate term merging may still remain. Differences among databases in bibliographic fields and citation indexing may also affect specific network indicators. Fifth, LDA modeling depends on text preprocessing, topic number, and hyperparameter settings, and some high-weight terms may contain publication-format or database noise. Clinical interpretation of topic labels is partly subjective; therefore, LDA topics should not be regarded as fixed disease classifications. Finally, bibliometric analysis describes the knowledge structure of published literature, cannot include unpublished negative results, and cannot establish causal relationships regarding disease mechanisms or treatment effects. The clinical interpretations and future directions proposed in this article still require validation through prospective cohorts and randomized controlled trials. Conclusion From a multidatabase perspective, this study systematically mapped the knowledge structure and thematic evolution of COPD-OSA overlap syndrome research. The knowledge base of this field has long centered on nocturnal hypoxemia, acute exacerbations, cardiovascular outcomes, and PAP therapy, and has recently expanded toward complex comorbidities, biomarkers, clinical phenotypes, patient function, and digital screening. Overall, the research focus is shifting from confirming disease coexistence toward identifying high-risk phenotypes and implementing stratified management, but mechanistic research and therapeutic evidence still mainly derive from observational, small-sample, or exploratory studies. Future studies should clarify the real-world benefits of PAP, noninvasive ventilation, and integrated treatment for different patient groups through harmonized case definitions, multicenter longitudinal cohorts, standardized core outcomes, and phenotype-stratified trials. The knowledge map generated in this study can be used to identify evidence gaps and research priorities, but it should not be interpreted as clinical causal evidence or a direct treatment recommendation. Funding Statement This work did not receive any specific financial grant. Generative AI Statement The authors declare that no Gen AI was used in the creation of this manuscript. Data Sharing Statement The bibliographic data analyzed in this study were retrieved from the Web of Science Core Collection, Scopus, and PubMed. The Web of Science Core Collection and Scopus are subscription-based bibliographic databases, whereas PubMed is publicly accessible. The complete search strategies, inclusion and exclusion criteria, and detailed parameter settings for VOSviewer, CiteSpace, and Bibliometrix are provided in the Supplementary Material. No primary experimental data were generated in this study. Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. All authors have reviewed and consented to the submission and publication of this manuscript. Disclosure The authors declare that they have no competing financial or non-financial interests related to this work. References Associated Data Data Availability Statement The bibliographic data analyzed in this study were retrieved from the Web of Science Core Collection, Scopus, and PubMed. The Web of Science Core Collection and Scopus are subscription-based bibliographic databases, whereas PubMed is publicly accessible. The complete search strategies, inclusion and exclusion criteria, and detailed parameter settings for VOSviewer, CiteSpace, and Bibliometrix are provided in the Supplementary Material. No primary experimental data were generated in this study.