Comparative effectiveness of triple versus dual inhaled therapy (ICS/LABA or LABA/LAMA) for chronic obstructive pulmonary disease: a systematic review and network meta-analysis Authors: Weimian Hu, Qinglin Li Published: 2026 ABSTRACT BackgroundThe optimal role of triple therapy (ICS/LAMA/LABA) versus dual therapy in COPD remains uncertain, particularly given the trade-off between modest efficacy gains and potential safety concerns. In particular, whether triple therapy provides additional benefit over dual regimens consisting of either ICS/LABA or LABA/LAMA remains an important clinical question. This study aimed to evaluate the comparative effectiveness and safety of triple versus dual inhaled therapies in COPD using a network meta-analysis.MethodsA systematic search of PubMed, Embase, Cochrane Library, CNKI, Wanfang, and VIP databases was conducted from inception to July 2023. Randomized controlled trials comparing triple inhaled therapy with either ICS/LABA or LABA/LAMA dual inhaled therapy in COPD were eligible. Outcomes included moderate-to-severe exacerbations, forced expiratory volume in one second (FEV₁), St. George’s Respiratory Questionnaire (SGRQ), Transition Dyspnea Index (TDI), and pneumonia incidence. Data were synthesized using a frequentist random-effects NMA. Treatment hierarchies were estimated using surface under the cumulative ranking curve (SUCRA).ResultsSeven RCTs (n = 9,968 participants) were included. Triple therapy did not consistently reduce exacerbation rates compared with dual therapy, while dual regimens such as FF/VI and BUD/FORM ranked favorably for exacerbation prevention (SUCRA 75.6% and 71.4%). Triple therapy and selected dual regimens improved lung function, though SUCRA rankings unexpectedly favored BUD/FORM. Improvements in SGRQ and TDI were modest and often below the minimal clinically important difference. Pneumonia risk tended to be higher with fluticasone-containing regimens, though differences did not reach statistical significance.ConclusionTriple therapy was associated with improvements in some outcomes, particularly lung function, but did not show consistent statistically significant superiority over dual therapy across all endpoints in this network meta-analysis. Given the small number of included trials and limited power for some comparisons, these findings should be interpreted cautiously and should not be taken as evidence of equivalence. Dual regimens remained effective options, while triple therapy may still be beneficial in selected high-risk patients. Personalized treatment based on exacerbation history, eosinophil count, and pneumonia risk remains essential.Trial registrationNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12890-026-04375-z. Introduction Chronic obstructive pulmonary disease (COPD) is a progressive, heterogeneous respiratory disorder characterized by airflow limitation, persistent symptoms, and recurrent exacerbations [1]. It is a leading cause of morbidity and mortality worldwide and imposes a substantial burden on healthcare systems and society [2]. Despite advances in pharmacological and non-pharmacological interventions, optimal management of COPD remains a challenge, particularly in patients with persistent symptoms and frequent exacerbations [3]. Inhaled pharmacotherapy forms the cornerstone of COPD management. Current guidelines recommend long-acting bronchodilators, including long-acting muscarinic antagonists (LAMAs) and long-acting β₂-agonists (LABAs), either alone or in combination, as first-line therapy for symptomatic patients. Inhaled corticosteroids (ICS) are often added in combination with LABA and/or LAMA for patients with recurrent exacerbations, especially those with elevated blood eosinophil counts [4]. These therapeutic strategies have been refined through pivotal clinical trials, yet the comparative effectiveness and safety of dual versus triple inhaled therapy remain subjects of ongoing debate. Triple therapy (ICS/LAMA/LABA) is hypothesized to provide additive benefits by targeting multiple pathophysiological pathways, potentially improving lung function, reducing exacerbations, and enhancing health-related quality of life. Large randomized controlled trials (RCTs), such as the IMPACT and ETHOS studies [5, 6], have reported modest reductions in exacerbation rates with triple therapy compared with dual therapy, but also highlighted an increased risk of pneumonia, particularly with fluticasone-containing regimens. Dual therapies, particularly combinations such as fluticasone furoate/vilanterol (FF/VI) and budesonide/formoterol (BUD/FORM), remain widely prescribed and are often sufficient for many patients [7]. Network meta-analysis (NMA) offers a robust methodological framework to compare multiple interventions simultaneously, integrating both direct and indirect evidence to establish a clinically interpretable hierarchy of treatments [8]. Previous meta-analyses have provided insights into the benefits and risks of triple versus dual therapy, but results have been inconsistent, and the relative ranking of different inhaled regimens across multiple outcomes, including exacerbation prevention, lung function, patient-reported quality of life, and adverse events, remains unclear. Therefore, we conducted a systematic review and NMA of RCTs to comprehensively evaluate the comparative effectiveness and safety of triple versus dual inhaled therapies in COPD. Our objective was to clarify the relative benefits and harms across key clinical outcomes, thereby informing individualized treatment strategies and optimizing evidence-based care for COPD patients. Methods Search strategy A comprehensive and systematic search of both international and Chinese databases was conducted to identify all relevant studies. The databases included PubMed, Embase, the Cochrane Library, CNKI, Wanfang Data, and VIP. Given the long-term nature of pharmacological interventions for COPD, no restrictions were imposed on the publication year, but the search was limited to articles published in English or Chinese. The English search strategy combined the following terms using Boolean operators: “triple inhalation therapy” OR “triple therapy” AND “dual inhalation therapy” OR “dual therapy” AND “chronic obstructive pulmonary disease” OR “COPD”. The Chinese-language search strategy used the terms (“三联吸入剂” OR “双联吸入剂”) AND “慢性阻塞性肺疾病”. In addition to the database searches, reference lists of all included studies and relevant review articles were manually screened to identify any additional eligible studies. No separate grey literature searches were performed. The search strategy was designed and conducted in accordance with the PRISMA guidelines for systematic reviews and network meta-analyses. Inclusion and exclusion criteria Studies were considered eligible if they met the following criteria: (1) Study design: randomized controlled trials (RCTs) or controlled clinical studies that directly compared triple inhaled therapy with dual inhaled therapy; (2) Population: adult patients diagnosed with COPD, regardless of sex, ethnicity, or geographic location; (3) Interventions: any formulation of triple inhaled therapy compared with dual inhaled therapy; (4) Outcomes: studies reporting at least one of the following endpoints moderate-to-severe exacerbation rate, forced expiratory volume in one second (FEV₁), St. George’s Respiratory Questionnaire (SGRQ) score [9], Transition Dyspnea Index (TDI) score, or incidence of pneumonia; and (5) Language: publications in English or Chinese. Exclusion criteria were: (1) studies with a sample size fewer than 30 participants; (2) reviews, protocols, case reports, conference abstracts, or other non-original research; (3) duplicate publications or overlapping datasets; (4) studies with insufficient outcome data or that where essential information could not be extracted; and (5) poor-quality studies with serious methodological flaws or incomplete reporting that precluded reliable interpretation. Screening and data extraction Two reviewers independently screened the titles and abstracts of all retrieved records to exclude obviously irrelevant articles. Full texts of potentially eligible studies were then reviewed in detail to confirm inclusion based on the pre-specified criteria. Any discrepancies between the two reviewers were resolved through discussion, with a third reviewer consulted when consensus could not be reached. Data extraction was also performed independently by two reviewers using a standardized form. Extracted information included: (1) study characteristics included first author, year of publication, country, design, and sample size; (2) patient demographics and baseline characteristics, such as age, sex distribution, disease severity, and lung function; (3) details of interventions and comparators, like types and dosage of triple versus dual inhaled therapy; and (4) outcomes of interest, including exacerbation rates, FEV1, SGRQ, TDI, and pneumonia incidence. When multiple publications reported data from the same cohort, the most recent or most complete dataset was included. Attempts were made to contact study authors when key data were missing. All extracted data were cross-checked for accuracy before analysis. Quality assessment The methodological quality of included randomized controlled trials was evaluated using the Cochrane Risk of Bias Tool (RoB 2.0). Each study was assessed across five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of outcomes, and selection of reported results. Each domain was rated as “low risk,” “some concerns,” or “high risk” of bias, and an overall risk-of-bias judgment was assigned accordingly. Quality assessments were carried out independently by two reviewers, and disagreements were resolved by consensus or arbitration by a third reviewer. The results of the risk-of-bias assessments were presented in both tabular and graphical formats. In addition, sensitivity analyses were pre-specified to evaluate the robustness of the findings by excluding studies judged to be at high risk of bias. Consistency assessment Consistency, defined as the agreement between direct and indirect evidence across treatment comparisons, is a fundamental assumption of network meta-analysis (NMA). To assess consistency, we employed both local and global approaches. Locally, a node-splitting method was applied to compare direct and indirect estimates within each closed loop of the treatment network. Globally, a design-by-treatment interaction model was used to evaluate inconsistency across the entire network. A P value < 0.05 was considered indicative of significant inconsistency. In the absence of inconsistency, results from the consistency model were reported as the primary estimates. SUCRA assessment The relative ranking probability of each intervention was estimated using the Surface Under the Cumulative Ranking curve (SUCRA). SUCRA values range from 0% to 100%, where 100% indicates that the treatment is certain to be the most effective and 0% indicates certainty of being the least effective. Rankograms and SUCRA plots were generated to visualize the probability distribution of treatment rankings across different outcomes. Additionally, the mean rank and probability of being the best treatment (PrBest) were calculated to provide complementary insights. This approach allows for a clinically interpretable hierarchy of interventions while considering both magnitude and uncertainty of effect estimates. Statistical analysis All pairwise and network meta-analyses were performed within a frequentist framework using a random-effects model to account for potential heterogeneity across studies. Moderate-to-severe exacerbations and pneumonia were analyzed as dichotomous outcomes using risk ratios (RRs) with 95% confidence intervals (CIs), whereas FEV₁, SGRQ, and TDI were analyzed as continuous outcomes using mean differences (MDs) with 95% CIs. Between-study heterogeneity was quantified using the τ² statistic and the I² statistic, with thresholds of 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively. Consistency and SUCRA analyses were conducted using the “netmeta” package in R as previous study [10], while visualization of forest plots, rankograms, and funnel plots was performed using Stata (version 19) and R (version 4.4.2). To improve interpretability, forest plots were organized into clinically relevant comparison categories, including triple therapy versus ICS/LABA, triple therapy versus LABA/LAMA, and ICS/LABA versus LABA/LAMA. Statistical significance was set at a two-sided P < 0.05. Protocol and registration This systematic review and network meta-analysis was not prospectively registered, and no separate protocol publication is available. This should be considered a methodological limitation. However, the review was conducted using predefined eligibility criteria, outcomes, and analytical methods, and was reported in accordance with PRISMA guidance. Results Study Characteristics The initial search across six databases yielded 2,135 records. After removing 380 duplicates and 25 records for other reasons, 1,730 records were screened by title and abstract. Of these, 1,590 were excluded as clearly irrelevant. The full text of 140 potentially eligible articles was reviewed in detail, resulting in the exclusion of 133 studies. The main reasons for exclusion were interventions or comparators not meeting inclusion criteria (n = 85), insufficient or incomplete outcome data (n = 28), and poor methodological quality (n = 20). Ultimately, 7 studies were included in the qualitative and quantitative synthesis. The study selection process is summarized in the PRISMA 2020 flow diagram (Fig. 1). Fig. 1PRISMA 2020 flow diagram PRISMA 2020 flow diagram The main characteristics of the included studies are presented in Table 1. All seven studies were randomized controlled trials published between 2017 and 2020. The total sample sizes ranged from 72 to 4,151 participants, with individual trial sizes varying substantially. Table 1Characteristics of included studiesFirst authorSample sizePublished yearInterventionsStudy groupControl groupStudy groupControl groupLipson DA [38]2070 / 413441512018UMEC-VI / FF-VIFF-UMEC-VIFerguson GT [39]639 / 6253142019BUD-GLY-FORM / GLY-FORBUD-FORMLipson DA [11]9118992017FF-UMEC-VIBUD-FORMWang C [12]144 / 144722020BUD-GLY-FORM / GLY-FORMBUD-FORMIchinose M [40]139 / 138702019BUD-GLY-FORM / GLY-FORMBUD-FORMKato M [5]4151 / 413420702019FF-UMEC-VI / FF-VIUMEC-VIZheng J [13]8798702018FF-UMEC-VIBUD-FORM Table 1Characteristics of included studiesFirst authorSample sizePublished yearInterventionsStudy groupControl groupStudy groupControl groupLipson DA [38]2070 / 413441512018UMEC-VI / FF-VIFF-UMEC-VIFerguson GT [39]639 / 6253142019BUD-GLY-FORM / GLY-FORBUD-FORMLipson DA [11]9118992017FF-UMEC-VIBUD-FORMWang C [12]144 / 144722020BUD-GLY-FORM / GLY-FORMBUD-FORMIchinose M [40]139 / 138702019BUD-GLY-FORM / GLY-FORMBUD-FORMKato M [5]4151 / 413420702019FF-UMEC-VI / FF-VIUMEC-VIZheng J [13]8798702018FF-UMEC-VIBUD-FORM Characteristics of included studies Across all included studies, the primary outcomes were exacerbation rates, lung function assessed by FEV₁, patient-reported outcomes measured by SGRQ and TDI, and safety outcomes such as the incidence of pneumonia. Collectively, these studies provided sufficient data to allow for both pairwise meta-analyses and network meta-analysis of triple versus dual therapy in COPD. Risk of bias All seven trials adequately reported random sequence generation and allocation concealment, and thus were considered low risk for selection bias, as summarized in Fig. 2A. Similarly, blinding of participants, personnel, and outcome assessors was well described in nearly all studies, minimizing the risk of performance and detection bias. However, two studies (Lipson 2017 and Wang 2020) [11, 12] demonstrated a high risk of bias in incomplete outcome data (attrition bias), mainly due to differential dropout rates or incomplete follow-up reporting, as shown in Fig. 2B. In addition, one study (Zheng 2018) [13] was judged to have an unclear risk of reporting bias, as prespecified outcomes were insufficiently detailed. Fig. 2Risk of bias assessment of included randomized controlled trials. A Overall summary of risk of bias across all included studies, presented as percentages for each domain of the Cochrane Risk of Bias Tool. Green indicates low risk, yellow indicates unclear risk, and red indicates high risk. B Risk of bias assessment for individual studies. Each row represents a trial, and each column represents a bias domain. Most studies demonstrated low risk of bias, with isolated concerns related to incomplete outcome data and selective reporting Fig. 3Network meta-analysis of moderate-to-severe exacerbations. A Triple therapy vs. ICS/LABA. Forest plot showing mean differences and 95% confidence intervals (CIs) for comparisons between triple therapy regimens (ICS/LAMA/LABA) and ICS/LABA dual therapy. B Triple therapy vs. LABA/LAMA. Forest plot showing mean differences and 95% CIs for comparisons between triple therapy and LABA/LAMA dual bronchodilator therapy. C ICS/LABA vs. LABA/LAMA. Forest plot present risk ratios (RRs) with 95% confidence intervals (CIs). Squares represent point estimates, and horizontal lines indicate CIs. The vertical line represents the line of no effect (RR = 1) Risk of bias assessment of included randomized controlled trials. A Overall summary of risk of bias across all included studies, presented as percentages for each domain of the Cochrane Risk of Bias Tool. Green indicates low risk, yellow indicates unclear risk, and red indicates high risk. B Risk of bias assessment for individual studies. Each row represents a trial, and each column represents a bias domain. Most studies demonstrated low risk of bias, with isolated concerns related to incomplete outcome data and selective reporting Network meta-analysis of moderate-to-severe exacerbations. A Triple therapy vs. ICS/LABA. Forest plot showing mean differences and 95% confidence intervals (CIs) for comparisons between triple therapy regimens (ICS/LAMA/LABA) and ICS/LABA dual therapy. B Triple therapy vs. LABA/LAMA. Forest plot showing mean differences and 95% CIs for comparisons between triple therapy and LABA/LAMA dual bronchodilator therapy. C ICS/LABA vs. LABA/LAMA. Forest plot present risk ratios (RRs) with 95% confidence intervals (CIs). Squares represent point estimates, and horizontal lines indicate CIs. The vertical line represents the line of no effect (RR = 1) Fig. 4Network meta-analysis of pneumonia risk among COPD patients. Figure 4A–C. Forest plots showing the results of the network meta-analysis for moderate-to-severe exacerbations. Comparisons are organized into clinically relevant categories: (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Effect estimates are presented as risk ratios (RRs) with 95% confidence intervals (CIs). Squares represent point estimates, and horizontal lines indicate corresponding CIs. The vertical line represents the line of no effect (RR = 1). To improve readability, confidence interval lines were visually truncated where necessary to avoid overlap with text labels and numeric values Network meta-analysis of pneumonia risk among COPD patients. Figure 4A–C. Forest plots showing the results of the network meta-analysis for moderate-to-severe exacerbations. Comparisons are organized into clinically relevant categories: (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Effect estimates are presented as risk ratios (RRs) with 95% confidence intervals (CIs). Squares represent point estimates, and horizontal lines indicate corresponding CIs. The vertical line represents the line of no effect (RR = 1). To improve readability, confidence interval lines were visually truncated where necessary to avoid overlap with text labels and numeric values When summarized across all domains, the pooled assessment indicated that more than 85% of domains were at low risk of bias, with only a minority of unclear or high-risk judgments (Fig. 2A). These results suggest that the overall quality of evidence in this meta-analysis is robust and reliable, although caution is warranted when interpreting data from trials with higher attrition. Moderate-severe exacerbations Seven randomized controlled trials involving 9,968 patients contributed data on moderate-to-severe exacerbations. Most pairwise and network comparisons did not show statistically significant differences, with confidence intervals crossing the null value. When triple therapy regimens were compared with dual therapy, fluticasone furoate/umeclidinium/vilanterol (FF/UMEC/VI) showed a trend toward fewer exacerbations compared with budesonide/formoterol (BUD/FORM), although the result did not reach statistical significance. Similarly, budesonide/glycopyrronium/formoterol (BUD/GLY/FORM) did not demonstrate superiority over BUD/FORM. Comparisons of triple therapy versus other dual combinations, such as glycopyrronium/formoterol (GLY/FOR) or umeclidinium/vilanterol (UMEC/VI), also failed to show significant differences. In contrast, indirect comparisons suggested that some dual regimens, particularly fluticasone furoate/vilanterol (FF/VI), ranked favorably relative to BUD/FORM, indicating comparable efficacy in exacerbation prevention. However, given the limited number of trials and the sparse network for some comparisons, these findings should be interpreted as inconclusive rather than evidence of equivalence between triple and dual therapy. Large individual trials have suggested benefits of triple therapy in patients at higher exacerbation risk, and our results should be interpreted in that clinical context. SUCRA-based ranking probabilities are presented in Figure S1 and table S1. The ranking analysis indicated that FF/VI (SUCRA = 75.6%) and BUD/FORM (SUCRA = 71.4%) had the highest probabilities of being the most effective treatments for reducing moderate-to-severe exacerbations. These were followed by UMEC/VI (SUCRA = 57.3%), while both triple regimens, FF/UMEC/VI (41.7%) and BUD/GLY/FORM (35.6%), ranked lower than expected. The lowest-ranking interventions were GLY/FOR (35.0%) and GLY/FORM (33.4%), reflecting a lower likelihood of being the optimal choice for exacerbation reduction. Taken together, the findings suggest that triple therapy did not demonstrate consistent statistically significant superiority over dual therapy in preventing moderate-to-severe exacerbations within this network; however, these results should be interpreted cautiously and not as proof of equivalence. Pneumonia Six randomized controlled trials reported the incidence of pneumonia as a safety outcome. Across all treatment comparisons, most confidence intervals crossed the line of no effect, suggesting no statistically significant differences in pneumonia risk between triple and dual inhaled regimens. When individual comparisons were examined, fluticasone furoate/umeclidinium/vilanterol (FF/UMEC/VI) and fluticasone furoate/vilanterol (FF/VI) appeared to be associated with numerically higher risks of pneumonia compared with other regimens; however, these associations did not achieve statistical significance. Conversely, non–ICS-containing dual therapies such as glycopyrronium/formoterol (GLY/FOR) and umeclidinium/vilanterol (UMEC/VI) tended to show lower pneumonia risks, though again with wide confidence intervals. Where reported in the network estimates, pneumonia comparisons are expressed as risk ratios (RRs) with 95% confidence intervals rather than mean differences. Because the number of events and studies was limited, these estimates remain imprecise. SUCRA-based ranking analysis (Figure S2 and table S2) indicated that GLY/FORM (SUCRA = 87.6%) had the highest probability of being the safest regimen with respect to pneumonia, followed by GLY/FOR (69.6%) and BUD/FORM (69.9%). In contrast, regimens containing fluticasone furoate, especially FF/UMEC/VI (SUCRA = 3.9%) and FF/VI (18.7%), ranked lowest, consistent with a higher relative probability of pneumonia events. Taken together, while no direct comparison reached statistical significance, the ranking probabilities suggest that dual bronchodilator combinations may be associated with a lower risk of pneumonia, whereas triple therapies and fluticasone furoate–containing regimens tended to rank less favorably in the safety analysis. FEV1 All seven included trials reported changes in trough FEV₁, allowing for both direct and indirect comparisons across regimens (Fig. 5). Several regimens were associated with greater improvements in lung function compared with BUD/FORM. Specifically, FF/UMEC/VI (mean difference 170.06 mL, 95% CI 153.33–186.79) and FF/VI (73.06 mL, 95% CI 54.29–91.83) both significantly improved FEV₁ compared with BUD/FORM. Similarly, BUD/GLY/FORM and UMEC/VI showed moderate improvements over BUD/FORM, with mean differences of 71.03 mL (95% CI 52.68–89.37) and 116.06 mL (95% CI 96.47–135.65), respectively. In contrast, some comparisons involving dual bronchodilator regimens (GLY/FOR, GLY/FORM) versus ICS/LABA combinations demonstrated inconsistent or non-significant differences. Fig. 5Network meta-analysis of FEV₁ outcomes in COPD. Figure 5A–C. Forest plots presenting the network meta-analysis results for trough FEV₁. Comparisons are grouped as (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Effect estimates are expressed as mean differences (MDs) with 95% confidence intervals (CIs). Squares indicate point estimates, and horizontal lines represent CIs. The vertical line corresponds to no difference (MD = 0). Confidence intervals were truncated when necessary to enhance figure clarity and prevent overlap with annotations Network meta-analysis of FEV₁ outcomes in COPD. Figure 5A–C. Forest plots presenting the network meta-analysis results for trough FEV₁. Comparisons are grouped as (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Effect estimates are expressed as mean differences (MDs) with 95% confidence intervals (CIs). Squares indicate point estimates, and horizontal lines represent CIs. The vertical line corresponds to no difference (MD = 0). Confidence intervals were truncated when necessary to enhance figure clarity and prevent overlap with annotations SUCRA-based ranking results are presented in Figure S3 and table S3. However, the treatment ranking for FEV₁ should be interpreted with caution because the SUCRA results appeared sensitive to network structure and coding direction. We therefore focused primarily on the effect estimates and confidence intervals rather than ranking probabilities alone. Taken together, these findings indicate that several regimens, including FF/UMEC/VI and FF/VI, showed greater FEV₁ improvement than BUD/FORM in pairwise network estimates, but the ranking analysis should be interpreted cautiously pending confirmation of effect direction and coding consistency. SGRQ Six trials reported health-related quality of life measured by the St. George’s Respiratory Questionnaire (SGRQ). The pooled network meta-analysis is presented in Fig. 6. In general, most treatment comparisons did not reach statistical significance, as the confidence intervals for many pairwise estimates included the null value. When evaluated against BUD/FORM, both triple regimens showed favorable but inconsistent effects. FF/UMEC/VI was associated with a non-significant reduction in SGRQ score compared with BUD/FORM (mean difference − 0.40, 95% CI − 1.91 to 1.11). Similarly, BUD/GLY/FORM showed no clear superiority over BUD/FORM (− 0.26, 95% CI − 2.85 to 2.33). Among dual regimens, FF/VI demonstrated a statistically significant reduction in SGRQ score compared with BUD/FORM (− 2.21, 95% CI − 3.12 to − 1.30), indicating modest improvement in patient-reported quality of life. Fig. 6Network meta-analysis of SGRQ outcomes in COPD. Figure 6A–C. Forest plots showing the network meta-analysis results for St. George’s Respiratory Questionnaire (SGRQ) scores. Comparisons are categorized as (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Results are reported as mean differences (MDs) with 95% confidence intervals (CIs). Negative values indicate improvement in quality of life. Squares represent point estimates, and horizontal lines denote CIs. The vertical line indicates no difference (MD = 0). Confidence interval lines were truncated where required to avoid overlap and improve interpretability Network meta-analysis of SGRQ outcomes in COPD. Figure 6A–C. Forest plots showing the network meta-analysis results for St. George’s Respiratory Questionnaire (SGRQ) scores. Comparisons are categorized as (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Results are reported as mean differences (MDs) with 95% confidence intervals (CIs). Negative values indicate improvement in quality of life. Squares represent point estimates, and horizontal lines denote CIs. The vertical line indicates no difference (MD = 0). Confidence interval lines were truncated where required to avoid overlap and improve interpretability SUCRA-based ranking analysis (Figure S4 and table S4) further refined these findings. FF/UMEC/VI (SUCRA = 99.6%) had the highest probability of being the most effective regimen for improving SGRQ, with a 97.6% chance of being ranked first. Other favorable regimens included FF/VI (59.4%) and UMEC/VI (59.2%), whereas GLY/FOR (15.6%) and GLY/FORM (17.4%) ranked lowest, indicating limited benefit in quality-of-life improvement. Taken together, although most pairwise comparisons did not achieve statistical significance, the ranking analysis strongly suggests that ICS-containing regimens, particularly FF/UMEC/VI and FF/VI, are more likely to yield meaningful improvements in patient-reported quality of life compared with dual bronchodilator combinations. Transition dyspnea index (TDI) The comparative efficacy of different inhaled regimens on dyspnea, as measured by the TDI, is presented in Fig. 7. Pairwise and network meta-analysis showed no statistically significant differences between most treatment regimens, with confidence intervals generally crossing the line of no effect. Compared with BUD-FORM, BUD-GLY-FORM showed a trend toward slight improvement in TDI (MD: − 0.19, 95% CI: − 0.04 to 0.43), but this effect was not statistically significant. Similarly, GLY-FOR and GLY-FORM did not demonstrate superiority over BUD-FORM (MD: 0.03, 95% CI: − 0.25 to 0.31; MD: − 0.04, 95% CI: − 0.35 to 0.27, respectively). Comparisons among dual bronchodilator combinations also showed no clear benefit, with narrow confidence intervals indicating robust estimation. Fig. 7Figure 7A–C. Forest plots presenting the network meta-analysis results for Transition Dyspnea Index (TDI) scores. Comparisons are grouped into (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Effect estimates are expressed as mean differences (MDs) with 95% confidence intervals (CIs). Positive values indicate improvement in dyspnea. Squares indicate point estimates, and horizontal lines represent CIs. The vertical line corresponds to no difference (MD = 0). Confidence intervals were truncated where necessary to ensure clarity and prevent overlap with text Figure 7A–C. Forest plots presenting the network meta-analysis results for Transition Dyspnea Index (TDI) scores. Comparisons are grouped into (A) triple therapy versus ICS/LABA, (B) triple therapy versus LABA/LAMA, and (C) ICS/LABA versus LABA/LAMA. Effect estimates are expressed as mean differences (MDs) with 95% confidence intervals (CIs). Positive values indicate improvement in dyspnea. Squares indicate point estimates, and horizontal lines represent CIs. The vertical line corresponds to no difference (MD = 0). Confidence intervals were truncated where necessary to ensure clarity and prevent overlap with text SUCRA rankings (Figure S5 and table S5) further illustrated the relative probability of each intervention being the most effective in improving dyspnea. GLY-FORM achieved the highest probability of being ranked best (SUCRA 73.0%, PrBest 48.8%, Mean Rank 1.8), followed by BUD-FORM (SUCRA 64.3%, PrBest 28.5%, Mean Rank 2.1) and GLY-FOR (SUCRA 56.6%, PrBest 22.7%, Mean Rank 2.3). In contrast, BUD-GLY-FORM had the lowest probability of benefit (SUCRA 6.2%, PrBest 0.1%, Mean Rank 3.8). These results indicate that dual combinations, particularly GLY-FORM and BUD-FORM, may provide the most favorable effects on dyspnea symptoms, while triple combinations did not confer additional advantage in this outcome. Network meta-analysis The geometry of the treatment network for all included randomized controlled trials is illustrated in Fig. 8. The network included seven inhaled regimens: BUD-FORM, BUD-GLY-FORM, FF-UMEC-VI, FF-VI, GLY-FOR, GLY-FORM, and UMEC-VI. The size of each node reflects the number of patients randomized to that treatment, whereas the thickness of connecting lines represents the number of head-to-head trials comparing two regimens. Fig. 8Network geometry of the included randomized controlled trials. The size of each node is proportional to the number of patients randomized to that treatment, while the thickness of each line corresponds to the number of trials directly comparing two regimens. BUD-FORM acted as the most common comparator, providing multiple direct links across the network Network geometry of the included randomized controlled trials. The size of each node is proportional to the number of patients randomized to that treatment, while the thickness of each line corresponds to the number of trials directly comparing two regimens. BUD-FORM acted as the most common comparator, providing multiple direct links across the network BUD-FORM served as the most frequently evaluated comparator, forming direct connections with all other regimens. FF-UMEC-VI and BUD-GLY-FORM also demonstrated multiple connections, highlighting their frequent inclusion in clinical trials. In contrast, GLY-FOR and UMEC-VI were supported by relatively fewer studies, as indicated by thinner connecting lines and smaller node sizes. The well-connected structure of the network permitted robust estimation of both direct and indirect comparisons, thereby enhancing the precision of treatment ranking through SUCRA analyses. The balanced distribution of interventions within the network further supports the validity of subsequent consistency and transitivity assumptions required for network meta-analysis. Robustness and reliability of the network meta-analysis The robustness of our network meta-analysis was comprehensively evaluated through consistency and publication bias assessments.The agreement between direct and indirect evidence was examined using node-splitting and design-by-treatment interaction models. As shown in Fig. 9, no significant inconsistency was observed across treatment comparisons for all clinical endpoints, including exacerbations, pneumonia, FEV1, SGRQ, and TDI (all P > 0.05). The effect estimates derived from direct and indirect comparisons were highly concordant, supporting the validity of the transitivity and consistency assumptions underlying the network meta-analysis. Fig. 9Consistency assessment in the network meta-analysis across all outcomes. Forest plots display direct estimates, indirect estimates, and pooled results for each treatment comparison. Non-significant results from global and local inconsistency tests (p > 0.05) suggest overall agreement between direct and indirect evidence, supporting the validity of the network Consistency assessment in the network meta-analysis across all outcomes. Forest plots display direct estimates, indirect estimates, and pooled results for each treatment comparison. Non-significant results from global and local inconsistency tests (p > 0.05) suggest overall agreement between direct and indirect evidence, supporting the validity of the network Funnel plot analyses were performed for each outcome to evaluate the presence of small-study effects. As illustrated in Figures S6–S10, the plots for moderate-to-severe exacerbations, pneumonia, FEV1, SGRQ, and TDI were generally symmetrical, with most studies clustered around the pooled effect size. Although minor asymmetry was observed in a few outcomes, particularly in low-precision studies, no systematic bias or strong evidence of selective publication was detected. These findings suggest a low risk of publication bias across the included studies. Taken together, the absence of major inconsistency between direct and indirect estimates, combined with symmetrical funnel plots, provides strong evidence for the reliability and robustness of the network meta-analysis results. This strengthens confidence in the comparative effectiveness estimates derived for triple and dual inhaled therapies in COPD. Discussion Summary of evidence This network meta-analysis suggests that triple therapy may improve some outcomes, particularly lung function, but did not demonstrate consistent statistically significant superiority over dual therapy across all endpoints in this dataset. Because only seven trials were included, several comparisons were imprecise, and absence of statistical significance should not be interpreted as equivalence. Our findings therefore support a cautious interpretation: dual therapy remains an effective option for many patients, while triple therapy may still provide additional benefit in selected populations, especially those at higher exacerbation risk. Disagreements and agreements with other studies Triple therapy does not consistently reduce moderate-to-severe exacerbations compared to dual therapy in COPD patients, and certain dual regimens like fluticasone furoate/vilanterol (FF/VI) and budesonide/formoterol (BUD/FORM) have shown favorable rankings for exacerbation prevention and quality-of-life outcomes [6, 14]. While some triple regimens demonstrated improved lung function, these benefits were not uniformly reflected in SUCRA rankings, and ICS-containing regimens, particularly those with fluticasone furoate, were linked to a higher probability of pneumonia [6, 14]. These findings should be interpreted cautiously, as the limited number of included studies and imprecision in several comparisons reduce confidence in firm treatment hierarchies. Rather than indicating equivalence between triple and dual therapy, the present analysis suggests that any incremental benefit of triple therapy may depend on patient selection and clinical context. The current analysis aligns with and expands upon findings from pivotal RCTs concerning the efficacy and safety of inhaled therapies in COPD [14]. For instance, the IMPACT trial reported a 15% reduction in exacerbations with fluticasone furoate/umeclidinium/vilanterol (FF/UMEC/VI) compared to FF/VI, but this was accompanied by a 44% increased risk of pneumonia [15]. Similarly, the ETHOS trial indicated a dose-dependent reduction in exacerbations with budesonide-containing triple therapy, despite persistent elevated pneumonia incidence. These large-scale studies have influenced current guideline recommendations, yet they also highlighted the modest absolute benefit of triple therapy in unselected populations [16]. Our meta-analysis is broadly consistent with these data in showing that triple therapy may offer benefit in some settings, but the present network does not support an overly broad conclusion that triple therapy is uniformly superior or uniformly unnecessary. Instead, the results support a more individualized view in which triple therapy may be particularly relevant for patients with frequent exacerbations or other features associated with higher expected benefit. Our findings should not be interpreted as support for routine use of ICS/LABA as a preferred regimen in contemporary COPD management. Current GOLD recommendations no longer encourage LABA/ICS as standard therapy; instead, dual bronchodilation (LABA/LAMA) is generally preferred, with escalation to triple therapy (LABA/LAMA/ICS) when clinically indicated, particularly when an ICS indication is present [17]. Therefore, although ICS/LABA regimens were included in our network meta-analysis and some showed comparable efficacy in certain outcomes, these results reflect the historical trial evidence rather than current treatment recommendations [18, 19]. We consider that the role of ICS/LABA in COPD is now limited to specific clinical scenarios, such as patients with concomitant asthma or selected individuals already receiving ICS/LABA who demonstrate clinical benefit and require individualized reassessment. For most patients, treatment decisions should align with current guideline-based approaches incorporating exacerbation history, blood eosinophil count, and pneumonia risk [17, 20]. Nuanced observations on treatment outcomes The apparent superiority of BUD/FORM in SUCRA rankings for FEV₁ challenges the assumption that triple therapy consistently provides the greatest lung function benefit. This discrepancy could be attributed to heterogeneity in study populations, as trials involving FF/UMEC/VI often enrolled patients with more severe airflow limitation, potentially magnifying observed improvements but reducing comparability across studies [21]. Furthermore, FEV₁, while an objective measure, does not always correlate with patient-reported outcomes or exacerbation risk, raising questions about its sole use as a marker of treatment success [6, 22]. In addition, the discordance between some pairwise effect estimates and SUCRA rankings indicates that treatment rankings should not be interpreted in isolation. Where rankings and effect estimates appeared inconsistent, greater weight should be placed on the comparative effect sizes and their confidence intervals rather than the ranking probabilities alone. Quality-of-life outcomes, such as those assessed by the SGRQ and TDI, showed minimal differentiation among regimens, with most improvements falling below the minimal clinically important difference [6]. This observation is consistent with previous meta-analyses suggesting that pharmacological intensification alone may not fully address the multifactorial burden of COPD symptoms [23]. Dyspnea, fatigue, and reduced exercise capacity are influenced by systemic inflammation, comorbidities, and deconditioning, reinforcing the necessity of multimodal management, including pulmonary rehabilitation, smoking cessation, and optimized comorbidity care [22, 24]. Safety considerations The safety signal for pneumonia remains a critical consideration with ICS-containing regimens [14, 25]. The biological plausibility is well-established, as ICS can suppress airway immunity, impair macrophage function, and increase bacterial colonization [22]. Observational studies have consistently confirmed this risk, particularly with fluticasone furoate, which may be more potent yet also more immunosuppressive than budesonide [22, 26, 27]. Although this analysis did not demonstrate statistically significant differences in pneumonia risk, possibly due to limited statistical power, SUCRA rankings consistently indicated a higher risk for fluticasone-containing regimens [14, 25]. This supports a cautious approach to ICS use, especially in patients with recurrent infections, bronchiectasis, or low eosinophil counts [22]. Long-term use of ICS is also associated with other side effects, including oral candidiasis, dysphonia, mycobacterial disease, diabetes, and fractures [22, 25, 28, 29]. A systematic review and meta-analysis of direct-comparison studies found a clinically important intraclass difference in pneumonia risk between fluticasone and budesonide in COPD patients [27]. Fluticasone users had a 13.5% increased risk of pneumonia compared to budesonide users (summary relative risk RR ratio of 1.13; 95% CI: 1.09–1.19) [27]. For serious pneumonia, the risk was 14.4% higher among fluticasone users (pooled RR: 1.14; 95% CI: 1.09–1.20) [27]. This suggests that the type of ICS can influence pneumonia incidence [27]. The evolving role of blood eosinophils as a biomarker deserves emphasis [22]. Evidence indicates that patients with elevated eosinophil counts derive greater benefit from ICS-containing therapy in terms of exacerbation reduction [22]. Conversely, patients with persistently low eosinophil counts may experience little benefit from ICS while remaining exposed to the risk of pneumonia [22]. Although this analysis could not stratify by eosinophil count due to data limitations, this biomarker-driven approach is increasingly central to precision medicine in COPD and should inform future trial designs [22]. This analysis also sheds light on ongoing therapeutic controversies [30]. Some clinicians advocate for early escalation to triple therapy for patients with a high symptom burden, while others caution against overuse in stable patients without frequent exacerbations [22]. Our findings are more consistent with a selective rather than universal escalation strategy. Dual therapy appears appropriate for many patients, whereas escalation to triple therapy should be considered in the context of exacerbation history, eosinophil count, and pneumonia risk. This perspective aligns with the most recent Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines, which no longer recommend universal escalation to triple therapy but instead emphasize individualized care pathways [31, 32]. Strengths and limitations This study possesses several strengths, including the comprehensive inclusion of trials from both Western and Asian cohorts, the application of rigorous methods for consistency assessment, and the use of SUCRA rankings to provide clinically meaningful hierarchies [33–35]. First, the network was based on only seven trials, limiting statistical power and the stability of indirect comparisons and treatment rankings. Second, SUCRA values should be interpreted cautiously because rankings can be sensitive to sparse data and coding direction, particularly when absolute between-treatment differences are small. Nevertheless, limitations also exist, such as the relatively small number of included studies, which limits precision, especially for less commonly studied regimens like glycopyrrolate/formoterol (GLY/FOR) and umeclidinium/vilanterol (UMEC/VI) [33]. Variability in trial duration (ranging from 24 to 52 weeks) and outcome definitions may introduce heterogeneity. While industry sponsorship of most included trials raises the possibility of subtle bias, our risk-of-bias assessment and funnel plots did not suggest major distortion [36]. Lastly, subgroup analyses by eosinophil count, smoking status, or comorbidity profile were not feasible due to data limitations, but remain crucial for identifying patient subgroups that might benefit most from therapy intensification [22]. Future research directions Future research should address these identified gaps. Pragmatic RCTs and real-world evidence studies are needed to confirm whether the modest benefits of triple therapy observed in controlled settings translate into everyday clinical practice. Trials stratified by biomarkers such as eosinophils, microbiome profiles, or imaging phenotypes could further refine treatment personalization. Longer-term studies are also necessary to clarify the balance of benefits and harms, particularly the cumulative risks of pneumonia, osteoporosis, and metabolic complications associated with ICS use [37]. Finally, patient-centered outcomes, including dyspnea burden, daily activity, and health-related quality of life, should be prioritized as primary rather than secondary endpoints in future investigations to ensure that therapeutic advancements truly improve patients’ lives [6]. Conclusion In this network meta-analysis, triple therapy did not show consistent statistically significant superiority over dual therapy across all outcomes, although benefits in selected endpoints and patient groups remain possible. Given the limited number of included trials, these findings should be interpreted cautiously and not as proof of equivalence between treatment strategies. Dual regimens remain appropriate for many patients, whereas triple therapy may be most relevant for individuals with frequent exacerbations or other features suggesting likely benefit from ICS-containing treatment. Personalized treatment decisions should therefore integrate exacerbation history, eosinophil count, pneumonia risk, comorbidity profile, and patient preferences. Importantly, these findings should not be interpreted as supporting routine use of ICS/LABA in COPD, as current guideline recommendations favor LABA/LAMA or triple therapy depending on patient characteristics. Further well-powered studies are needed to clarify which subgroups derive the greatest net benefit from treatment escalation. Supplementary Information Supplementary Material 1 Supplementary Material 1