APPLICATION OF PROPENSITY SCORE MATCHING AND BAYESIAN HIERARCHICAL DESIGN METHODS TO INTEGRATE SINGLE-ARM STUDIES INTO NETWORK META-ANALYSES (NMAS): Opportunities and Pitfalls Illustrated in a Case Study Assessing Ablation/Radiation Therapies in Lung Cancer
Virtual ISPOR 2020 | TUESDAY, May 19th, 2020 | 5:45-6:00 PM ET
OBJECTIVES: Network meta-analyses (NMAs) generally include direct comparative evidence from randomized controlled trials (RCTs) and/or comparative observational studies; however, comparative evidence is limited in many disease/treatment areas. The objective of this analysis was to discuss opportunities and pitfalls associated with incorporating single-arm studies into NMAs, illustrated in a case study assessing the effectiveness of ablation/radiation therapies in lung cancer.
METHODS: A systematic literature review was conducted to identify RCTs, comparative observational studies, and single-arm studies assessing ablation/radiation therapies among adults with lung cancer. The outcomes were local tumor recurrence, overall survival, and complications. First, Bayesian hierarchical NMAs using direct comparative studies, down-weighting lower quality evidence, were conducted. Second, simulated comparative studies were obtained by matching relevant single-arm studies using optimal 1:1 matching; propensity scores were estimated by fitting a logistic regression model that included age, sex, tumor type, tumor size, and average number of tumors as covariates. Third, Bayesian hierarchical NMAs using both comparative and simulated comparative studies, down-weighting lower quality evidence, were conducted.
RESULTS: One RCT, 10 comparative observational studies, and 147 single-arm studies were identified. Seven to 22 simulated comparative studies were incorporated within each NMA, depending on the outcome. The conclusions of the Bayesian hierarchical NMAs were aligned between analyses using comparative or comparative and simulated comparative studies; however, differences in effect estimate magnitudes (0% – 44%) and treatment rankings were sometimes observed. Limitations of this analysis included sub-optimal reporting of covariates among single-arm studies limiting the ability to sufficiently match for cross-study differences and poor matching where cross-study differences existed.
CONCLUSIONS: Thoughtful integration of single-arm studies in NMAs may offer opportunities to utilize all available evidence and be especially useful in disease/treatment areas with many single-arm studies and limited direct comparative evidence or incomplete evidence networks. However, studies should clearly state the methodological limitations and present results stratified by study design.
Fill out the form below and download the presentation. Schedule a meeting to talk with our experts.
Rana is involved in a variety of data analytics/evidence synthesis projects such as clinical trial data analyses, systematic/narrative literature reviews, meta-analyses, network meta-analyses, matching-adjusted indirect treatment comparisons, and propensity-score matching analyses, as well as…
Dr. Chris Cameron is a global thought leader in health economics and outcomes research with over a decade of experience. Prior to joining EVERSANA, Chris was a partner at Cornerstone Research Group Inc., and…
Nicole directs U.S. and global projects to support multi-stakeholder medical device adoption and reimbursement. Nicole has more than 20 years of experience in health economics and research methods within academic, industry, and consulting roles.…