Network Meta-Analysis: Methods and Applications for Comparative Effectiveness

Network meta-analysis is a statistical technique that allows researchers to compare multiple treatments simultaneously within a single analysis by combining direct and indirect evidence from a network of randomized controlled trials. This method addresses a common problem in clinical research where many competing interventions exist for a condition but head-to-head trials comparing all options are unavailable. For students, researchers, and life-science professionals, understanding network meta-analysis is essential for interpreting comparative effectiveness research and for conducting rigorous evidence syntheses that inform clinical decisions.

At a Glance

Network meta-analysis extends traditional pairwise meta-analysis by enabling comparisons of multiple interventions even when direct trial evidence is missing. The table below summarizes the key distinctions and practical considerations.

What Network Meta-Analysis Adds to Evidence Synthesis

Pairwise meta-analysis is an established statistical tool for synthesizing evidence from multiple trials, but it is informative only about the relative efficacy of two specific interventions. The usefulness of pairwise meta-analysis is thus limited in real-life medical practice, where many competing interventions may be available for a certain condition and studies informing some of the pairwise comparisons may be lacking. This commonly encountered scenario has led to the development of network meta-analysis 11.

Network meta-analysis facilitates simultaneous comparison of the efficacy or safety of multiple interventions that may not have been directly compared in a randomized trial. This approach helps researchers study important and previously unanswerable questions, which has contributed to a rapid rise in the number of network meta-analysis publications in the biomedical literature 8.

The practical value of network meta-analysis lies in its ability to inform clinical decision-making. Clinicians regularly need to choose among several treatment options, and network meta-analysis can assist in assessing the comparative effectiveness of different treatments commonly used in clinical practice. However, if proper caution is not taken in conducting and interpreting network meta-analysis, inferences might be biased 6.

Core Concepts and Terminology

Direct, Indirect, and Mixed Comparisons

Direct comparisons come from randomized trials that compare two interventions head-to-head. Indirect comparisons estimate the relative effect between two interventions through a common comparator. For example, if trials compare intervention A to placebo and intervention B to placebo, the relative effect of A versus B can be estimated indirectly through the placebo arm. Mixed treatment comparisons combine both direct and indirect evidence when both are available for a given pair of interventions 8.

Network Structure and Connectivity

A network is formed by the set of interventions and the trials that compare them. Each intervention is a node, and each trial comparison is an edge connecting two nodes. Network connectivity requires that all interventions in the network are linked through a chain of comparisons. Without connectivity, indirect estimates cannot be derived for all pairs of interventions 8.

Treatment Ranking and SUCRA

Network meta-analysis produces effect estimates for all pairwise comparisons and also allows ranking of interventions according to the probability that each is the best for a given outcome. The surface under the cumulative ranking curve provides a numerical summary of these ranking probabilities. Higher SUCRA values indicate a greater likelihood of being among the most effective interventions 8. For example, in a network meta-analysis of traditional Chinese non-pharmacological therapies for chemotherapy-related symptoms, manual acupoint therapy ranked highest for improving quality of life with a SUCRA of 63.4 percent, while auricular therapy showed the greatest benefit for fatigue with a SUCRA of 100 percent 14.

Key Assumptions in Network Meta-Analysis

Transitivity

Transitivity is the assumption that the relative effect between two interventions is consistent across the different comparisons in the network. This means that the indirect estimate for A versus B through comparator C is valid only if the trials comparing A to C and B to C are sufficiently similar in terms of patient characteristics, study design, and outcome definitions. Transitivity is a clinical and methodological assumption that must be assessed before conducting the analysis 8.

Consistency

Consistency, also referred to as coherence, is the statistical manifestation of transitivity. It requires that direct and indirect estimates for the same comparison agree within random variation. When direct and indirect evidence conflict beyond what would be expected by chance, inconsistency is present, and the network meta-analysis results may be unreliable 8. A meta-epidemiological study examining network meta-analyses of proximal humerus fracture treatments specifically evaluated transitivity, coherence, and reliability, underscoring the importance of these assumptions in surgical research 27.

Homogeneity

Homogeneity refers to the assumption that the treatment effect is similar across trials within each direct comparison. This is the same assumption required in pairwise meta-analysis. Heterogeneity, or variability in treatment effects across trials, can threaten the validity of both pairwise and network meta-analyses 8.

Network Connectivity

All interventions in the network must be connected through a chain of comparisons. If some interventions are disconnected from the main network, indirect evidence cannot be derived for those disconnected nodes, and the network meta-analysis cannot provide estimates involving them 8.

Steps for Conducting a Network Meta-Analysis

Step 1: Define the Research Question

The research question must specify the population, interventions, comparators, outcomes, and study designs of interest. Unique considerations apply when developing research questions for network meta-analysis because the question must accommodate multiple interventions and the possibility of indirect comparisons 6. The question should be structured to allow identification of all relevant trials, including those that compare any pair of interventions in the network.

Step 2: Conduct a Systematic Literature Search

A comprehensive literature search is essential to identify all relevant randomized controlled trials. The search should cover multiple databases to minimize the risk of missing eligible studies. In a network meta-analysis of oral antihypertensive treatment during pregnancy, the electronic search covered PubMed, Embase, and CENTRAL to identify randomized controlled trials reporting perinatal outcomes 7. The NCBI Literature Resources and PubMed provide access to the biomedical literature for this purpose 4, 5.

Step 3: Abstract Data

Data abstraction for network meta-analysis requires extracting information on study characteristics, participant demographics, interventions, comparators, outcomes, and effect estimates. The abstraction process must be conducted systematically and ideally by two independent reviewers to reduce errors 6.

Step 4: Assess Risk of Bias

Risk of bias assessment is a critical component of any systematic review and network meta-analysis. The Cochrane Risk of Bias tool is commonly used for randomized controlled trials. In the antihypertensive treatment network meta-analysis, quality assessment was performed using the Cochrane Risk of Bias tool, and trustworthiness was assessed with the Trustworthiness in Randomised Controlled Trials Checklist 7. The EQUATOR Network provides reporting guidelines and resources that support transparent and complete reporting of research studies 2.

Step 5: Assess Network Geometry and Assumptions

Before fitting the statistical model, researchers should examine the network structure using graphical tools. Network plots display the interventions as nodes and the comparisons as edges, with edge thickness proportional to the number of trials. These plots help identify disconnected nodes, sparse comparisons, and potential sources of inconsistency 20.

Step 6: Fit the Network Meta-Analysis Model

Statistical models for network meta-analysis can be fitted using frequentist or Bayesian approaches. Both approaches combine direct and indirect evidence to produce mixed treatment effect estimates. The choice of approach depends on the research question, the complexity of the network, and the statistical expertise of the research team 8.

Step 7: Assess Consistency and Heterogeneity

Statistical tests and graphical methods can detect inconsistency between direct and indirect evidence. If inconsistency is present, researchers must investigate potential sources, such as differences in study populations, intervention doses, or outcome definitions. Heterogeneity within comparisons should also be quantified and explored 8.

Step 8: Present Results

Results of network meta-analysis should be presented using forest plots for pairwise comparisons, league tables showing all pairwise effect estimates, and ranking probabilities. Graphical tools can make the results accessible to non-statisticians and facilitate interpretation 20.

Step 9: Assess Confidence in the Evidence

Confidence in the results of a network meta-analysis should be evaluated systematically. The Confidence in Network Meta-Analysis framework considers six domains: within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence. This framework improves transparency and avoids the selective use of evidence when forming judgments 12.

Workflow Diagram for Conducting a Network Meta-Analysis

The following workflow outlines the sequential steps from question formulation to final reporting:

1. Formulate the research question with explicit specification of population, interventions, comparators, outcomes, and study design
2. Register the protocol in a prospective registry such as PROSPERO
3. Conduct a systematic search across multiple databases
4. Screen titles and abstracts against eligibility criteria
5. Retrieve full texts and assess eligibility
6. Extract data on study characteristics and outcomes
7. Assess risk of bias in included studies
8. Construct the network diagram and examine connectivity
9. Assess transitivity by comparing clinical and methodological characteristics across comparisons
10. Fit the network meta-analysis model using appropriate statistical methods
11. Evaluate consistency between direct and indirect evidence
12. Assess heterogeneity and explore potential sources
13. Perform sensitivity analyses to test robustness
14. Evaluate confidence in the evidence using frameworks such as CINeMA
15. Present results with appropriate graphical displays and interpret findings in the context of limitations

Applications in Life Sciences

Example 1: Antihypertensive Treatment During Pregnancy

A systematic review and network meta-analysis evaluated the effects of methyldopa, labetalol, and nifedipine for hypertensive disorders of pregnancy. The analysis included 23 trials with 3,989 women and found that compared to placebo or no treatment, labetalol and methyldopa significantly reduced the incidence of severe hypertension. Labetalol compared to nifedipine was associated with a reduction in preeclampsia and preterm birth. No significant differences were detected for other outcomes 7.

Example 2: Benzodiazepines for Generalized Anxiety Disorder

A network meta-analysis of 56 studies with 7,556 participants compared individual benzodiazepines for generalized anxiety disorder. All benzodiazepines were significantly better than placebo for efficacy, but no significant differences were observed between the different benzodiazepines. For treatment tolerability, diazepam showed a higher risk of discontinuation compared to placebo. The certainty of evidence varied from high to very low across comparisons 10.

Example 3: Exercise Modalities for Depressive Symptoms in Children and Adolescents

A Bayesian network meta-analysis of 41 trials with 4,563 participants evaluated different exercise types for depressive symptoms. Mind-body exercise was associated with the largest estimated reduction in depressive symptoms, followed by resistance training and aerobic exercise. Subgroup analyses suggested relatively larger effects for aerobic exercise in clinical populations and mind-body exercise in subclinical and healthy populations 16.

Example 4: Digital Interventions for Smoking Cessation

A network meta-analysis of 152 randomized controlled trials assessed digital interventions for smoking cessation. Personalized interventions significantly improved smoking cessation rates compared with standard care, and text message-based interventions were the most effective among technology types. Intervention effectiveness was influenced by age, with middle-aged individuals benefitting more than younger individuals 21.

Example 5: Myopia Control in Children and Adolescents

A network meta-analysis of 16 randomized controlled trials with 9,084 participants compared sports and outdoor activities for myopia control. Racket sports ranked highest for slowing axial elongation, followed by increased outdoor time. For refractive outcomes, visual tracking exercise ranked highest for improving spherical equivalent. Subgroup analyses showed that increased outdoor time remained effective in children aged 8.5 years or younger and in interventions lasting more than 24 weeks 17.

Options and Tradeoffs in Network Meta-Analysis

Frequentist versus Bayesian Approaches

Frequentist network meta-analysis uses standard statistical methods to estimate treatment effects and confidence intervals. Bayesian approaches incorporate prior distributions and produce posterior probability distributions for treatment effects. Both approaches can produce valid results, but they differ in their assumptions, computational requirements, and interpretation. Bayesian methods naturally accommodate complex models and provide direct probability statements about treatment rankings 8.

Fixed-Effect versus Random-Effects Models

Fixed-effect models assume that the true treatment effect is the same across all trials. Random-effects models allow for between-trial variability in treatment effects. Random-effects models are generally preferred in network meta-analysis because clinical trials typically exhibit some degree of heterogeneity. The choice between these models should be guided by the degree of heterogeneity and the research question 8.

Component Network Meta-Analysis

Component network meta-analysis enables disentangling individual component effects from multicomponent treatments. This approach is useful when interventions are combinations of multiple components and researchers want to estimate the contribution of each component. A leave-one-out algorithm has been proposed to quantify the contribution of evidence from constituent comparisons to the disentangled component effect estimates 24.

Observations and Measurements in Network Meta-Analysis

Network Plots

Network plots provide a visual representation of the evidence base. Each intervention is represented as a node, and each comparison is represented as an edge connecting two nodes. The thickness of edges can represent the number of trials or participants for each comparison. Node size can represent the number of participants receiving each intervention. These plots help researchers and readers quickly assess the structure and completeness of the network 20.

Contribution Matrix

The percentage contribution matrix shows how much information each study contributes to the results from network meta-analysis. This matrix is key to judgments about within-study bias and indirectness. The contribution matrix can easily be computed using a freely available web application 12.

Prediction Intervals

Prediction intervals provide a range within which the true treatment effect in a new study is expected to fall. These intervals account for both the uncertainty in the pooled estimate and the between-study variability. Prediction intervals are particularly informative for clinical decision-making because they reflect the range of effects that might be observed in future patients 8.

Records and Documentation

Protocol Registration

Prospective registration of the systematic review and network meta-analysis protocol is recommended to promote transparency and reduce the risk of reporting bias. The PROSPERO registry is commonly used for this purpose. For example, the benzodiazepine network meta-analysis was registered with PROSPERO under registration number CRD42022330264 10, and the exercise and sleep disorder network meta-analysis was registered under CRD420251118367 15.

Data Extraction Forms

Standardized data extraction forms should capture study characteristics, participant demographics, intervention details, comparator details, outcome definitions, effect estimates, and measures of variability. These forms should be piloted and refined before full data extraction begins.

Statistical Analysis Code

Documenting the statistical analysis code is essential for reproducibility. The code should be made available alongside the published results to allow other researchers to verify the analyses and conduct additional sensitivity analyses.

Quality and Welfare Controls

Risk of Bias Assessment

Risk of bias assessment should be conducted for all included studies using validated tools. The Cochrane Risk of Bias tool for randomized controlled trials is widely used. In the antihypertensive treatment network meta-analysis, the overall quality of included studies was low to moderate 7. In the opioid-free anesthesia network meta-analysis, moderate to high risk of bias was identified in over 70 percent of studies reporting the primary outcome 22.

Certainty of Evidence Assessment

The certainty of evidence should be assessed using established frameworks. The Confidence in Network Meta-Analysis framework considers six domains and provides a structured approach to evaluating confidence in the results 12. The GRADE approach for network meta-analysis is also used. In the benzodiazepine network meta-analysis, the certainty of evidence varied from high to very low, with 40 comparisons scored as very low, 7 as low, and 814 as high 10.

Sensitivity Analyses

Sensitivity analyses should be conducted to test the robustness of the results to assumptions and decisions made during the review process. These analyses might exclude studies at high risk of bias, use different statistical models, or restrict the analysis to specific subgroups. In the myopia control network meta-analysis, sensitivity analyses supported the robustness of the findings 17.

Common Failure Patterns in Network Meta-Analysis

Failure to Assess Transitivity

Transitivity is a clinical assumption that must be assessed before conducting the statistical analysis. Failure to assess whether the trials in the network are sufficiently similar in terms of patient characteristics, study design, and outcome definitions can lead to biased indirect estimates. Researchers should compare the distribution of effect modifiers across the different comparisons in the network 8.

Ignoring Inconsistency

Inconsistency between direct and indirect evidence can invalidate network meta-analysis results. Researchers must test for inconsistency and investigate potential sources when it is detected. Ignoring inconsistency can lead to misleading conclusions about the relative effectiveness of interventions 8.

Overinterpreting Treatment Rankings

Treatment rankings based on SUCRA values are probabilistic statements, not definitive conclusions. Rankings can be sensitive to small differences in treatment effects and to the structure of the network. Researchers and readers should interpret rankings cautiously, particularly when confidence intervals for treatment effects overlap substantially 8.

Inadequate Reporting

Incomplete reporting of methods and results is a common problem in network meta-analysis. Key components of reporting include the proposed analysis methods, assessment of risk of bias within the included studies, reporting the overall quality of the available evidence, and defining the parameters in which the results will be presented 9. The EQUATOR Network provides reporting guidelines that support transparent and complete reporting 2.

Lack of Standardization

There is a lack of consistency and standardization in network meta-analysis methodology due to the rapidly evolving nature of the field. Researchers should follow guidance from key methodology groups, as these provide valuable tools for conducting and reporting network meta-analyses 9.

Limitations of Network Meta-Analysis

Complexity of Conduct and Interpretation

The conduct and interpretation of network meta-analyses are more complex than pairwise meta-analyses. There are additional model assumptions, including network connectivity, homogeneity, transitivity, and consistency, and additional outputs such as network plots and SUCRA values 8. This complexity can lead to errors if researchers lack appropriate statistical expertise.

Dependence on the Quality of Primary Studies

Network meta-analysis cannot overcome limitations in the primary studies. If the included trials have high risk of bias, the network meta-analysis results will also be biased. In the opioid-free anesthesia network meta-analysis, moderate to high risk of bias was identified in over 70 percent of studies reporting the primary outcome, limiting the certainty of the findings 22.

Potential for Bias

Network meta-analysis is susceptible to various sources of bias, including reporting bias, selective outcome reporting, and publication bias. The Confidence in Network Meta-Analysis framework explicitly considers reporting bias as one of its six domains 12.

Heterogeneity in Intervention Definitions

Inconsistent definitions of interventions across trials can complicate network meta-analysis. In the digital interventions for smoking cessation review, methodological heterogeneity, potential bias, and inconsistent definitions of numerical interventions were identified as limitations 21.

Limited Long-Term Follow-Up Data

Many trials have limited follow-up durations, which restricts the ability of network meta-analysis to assess long-term outcomes. The smoking cessation review noted that long-term follow-up data remain limited 21.

Safety and Regulatory Context

Evidence-Based Clinical Decision-Making

Network meta-analysis provides evidence that can inform clinical decision-making, but it does not replace clinical judgment. Clinicians must consider individual patient characteristics, preferences, and values when applying network meta-analysis results to specific patients. The results of network meta-analysis should be interpreted in the context of the overall evidence base and the certainty of the evidence 6.

Regulatory Considerations

Regulatory agencies may consider network meta-analysis evidence when evaluating the comparative effectiveness of interventions. However, the quality and certainty of the evidence must be carefully assessed. Researchers conducting network meta-analyses that may inform regulatory decisions should follow established methodological guidance and report their methods transparently 9.

Research Data Management

The Research Data Framework from the National Institute of Standards and Technology provides guidance on managing research data throughout the research lifecycle 1. Proper data management is essential for the reproducibility and transparency of network meta-analyses.

Experimental Design Considerations

The Experimental Design Assistant from the NC3Rs provides support for designing rigorous experiments, which is relevant to the planning of primary studies that will contribute to future evidence syntheses 3. Well-designed randomized controlled trials are the foundation of reliable network meta-analyses.

Professional Escalation Criteria

Researchers should seek additional statistical expertise or methodological consultation when encountering the following situations:

1. The network has disconnected nodes or sparse comparisons that limit the ability to estimate all pairwise effects
2. Statistical tests indicate significant inconsistency between direct and indirect evidence
3. The network meta-analysis model fails to converge or produces unstable estimates
4. There is substantial heterogeneity that cannot be explained by subgroup analyses or meta-regression
5. The research question involves complex interventions with multiple components that require component network meta-analysis
6. The results will inform regulatory decisions or clinical guidelines where the certainty of evidence must be rigorously evaluated

In these situations, consultation with a statistician specializing in evidence synthesis or a methodologist with expertise in network meta-analysis is recommended. The CINeMA framework and other structured approaches can help researchers systematically evaluate the confidence in their results 12.

Frequently Asked Questions

What is the difference between network meta-analysis and pairwise meta-analysis?

Pairwise meta-analysis compares two interventions based on head-to-head data from randomized trials. Network meta-analysis facilitates simultaneous comparison of the efficacy or safety of multiple interventions that may not have been directly compared in a randomized trial 8. Network meta-analysis combines direct and indirect evidence within a network of trials, allowing comparisons between interventions that have never been directly studied against each other 6.

When should network meta-analysis be used instead of pairwise meta-analysis?

Network meta-analysis should be used when there are multiple potential interventions for a condition and researchers need to compare and rank all available interventions. Pairwise meta-analysis is informative only about the relative efficacy of two specific interventions, which limits its usefulness in real-life medical practice where many competing interventions may be available 11. Network meta-analysis is particularly valuable when studies informing some of the pairwise comparisons are lacking 8.

What are the key assumptions of network meta-analysis?

The key assumptions are network connectivity, homogeneity, transitivity, and consistency. Network connectivity requires that all interventions are linked through a chain of comparisons. Homogeneity requires that treatment effects are similar across trials within each direct comparison. Transitivity requires that the relative effect between two interventions is consistent across the different comparisons in the network. Consistency is the statistical manifestation of transitivity, requiring that direct and indirect estimates agree within random variation 8.

How is treatment ranking determined in network meta-analysis?

Treatment ranking is determined using the surface under the cumulative ranking curve. SUCRA values provide a numerical summary of the probability that each intervention is among the most effective for a given outcome. Higher SUCRA values indicate a greater likelihood of being among the most effective interventions 8. Rankings should be interpreted cautiously, particularly when confidence intervals for treatment effects overlap substantially.

What is the Confidence in Network Meta-Analysis framework?

The Confidence in Network Meta-Analysis framework is a methodological approach to evaluate confidence in the results from network meta-analyses. It considers six domains: within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence. The framework improves transparency and avoids the selective use of evidence when forming judgments 12.

What are the common problems encountered when conducting network meta-analyses?

Common problems include disconnected networks, sparse comparisons, inconsistency between direct and indirect evidence, heterogeneity, and inadequate reporting. Researchers may also encounter difficulties with model convergence, particularly in complex networks. The conduct and interpretation of network meta-analyses are more complex than pairwise meta-analyses, and there are additional model assumptions and outputs to consider 8.

How should the results of a network meta-analysis be presented?

Results should be presented using forest plots for pairwise comparisons, league tables showing all pairwise effect estimates, and ranking probabilities. Graphical tools can make the results accessible to non-statisticians and facilitate interpretation 20. Key components of reporting include the proposed analysis methods, assessment of risk of bias within the included studies, reporting the overall quality of the available evidence, and defining the parameters in which the results will be presented 9.

What is component network meta-analysis?

Component network meta-analysis enables disentangling individual component effects from multicomponent treatments. This approach is useful when interventions are combinations of multiple components and researchers want to estimate the contribution of each component. A leave-one-out algorithm has been proposed to quantify the contribution of evidence from constituent comparisons to the disentangled component effect estimates 24.

Related Articles

• Statistical Analysis Methods
• Statistical Analysis Methods
• Statistical Analysis Methods
• Rna Sequencing Analysis Workflow
• Rna Sequencing Analysis Workflow

References and Further Reading

• Research Data Framework. National Institute of Standards and Technology.
• EQUATOR Network. EQUATOR Network.
• Experimental Design Assistant. NC3Rs.
• NCBI Literature Resources. National Center for Biotechnology Information.
• PubMed. National Library of Medicine.
• Network meta-analysis: an introduction for clinicians.. Internal and emergency medicine, 2017.
• Oral antihypertensive treatment during pregnancy: a systematic review and network meta-analysis.. American journal of obstetrics and gynecology, 2025.
• Network Meta-Analysis.. Methods in molecular biology (Clifton, N.J.), 2022.
• Understanding network meta-analysis methodology for the ophthalmologist.. Current opinion in ophthalmology, 2024.
• Comparative Efficacy and Safety of Benzodiazepines in the Treatment of Patients with Generalized Anxiety Disorder: A Systematic Review and Network Meta-Analysis.. Psychotherapy and psychosomatics, 2025.
• GetReal in network meta-analysis: a review of the methodology.. Research synthesis methods, 2016.
• CINeMA: An approach for assessing confidence in the results of a network meta-analysis.. PLoS medicine, 2020.
• Network meta-analysis: Looping back.. Research synthesis methods, 2024.
• Comparative effectiveness of traditional Chinese non-pharmacological therapies for chemotherapy-related symptoms in cancer patients: a systematic review and network meta-analysis.. 2026.
• Comparative effectiveness of single vs. combined exercise modalities on sleep disorders in Chinese adolescents: a systematic review and network meta-analysis.. 2026.
• Comparative effectiveness and dose-response relationships of exercise for depressive symptoms in children and adolescents: a Bayesian network meta-analysis.. 2026.
• Comparative effectiveness of outdoor and exercise interventions for myopia control in children and adolescents: a systematic review and network meta-analysis of randomized clinical trials.. 2026.
• Comparative Effectiveness of AI-Assisted Telerehabilitation, Telerehabilitation, In-Person Care, and Usual Care for Chronic Nonspecific Low Back Pain: Bayesian Network Meta-Analysis.. 2026.
• Comparative effectiveness of various exercise modalities on arterial stiffness and endothelial function in older adults: a systematic review and network meta-analysis of randomised controlled trials.. 2026.
• Graphical Tools for Network Meta-Analysis in STATA. PLoS ONE, 2013.
• Efficacy of digital interventions for smoking cessation by type and method: a systematic review and network meta-analysis. Nature Human Behaviour, 2025.
• Effectiveness and safety of opioid-free anesthesia compared to opioid-based anesthesia: a systematic review and network meta-analysis. Journal of Anesthesia Analgesia and Critical Care, 2025.
• Alpha-Synuclein Seed Amplification Assays in Parkinson’s Disease: A Systematic Review and Network Meta-Analysis. Clinics and Practice, 2025.
• A leave-one-out algorithm for contribution analysis in component network meta-analysis. BMC Medical Research Methodology, 2025.
• Methodologies for network meta-analysis of randomised controlled trials in pain, anaesthesia, and perioperative medicine: a narrative review. British Journal of Anaesthesia, 2025.
• Network meta-analysis in psychiatric research: Opportunities and caveats. Psychiatria Danubina, 2018.
• Transitivity, coherence, and reliability of network meta-analyses comparing proximal humerus fracture treatments: a meta-epidemiological study. BMC Musculoskeletal Disorders, 2024.

This article is educational and does not replace institutional policy, professional advice, or applicable safety and regulatory requirements.