OBJECTIVES: Lung cancer patients who are ineligible for surgery are treated with local treatments such as microwave ablation (MWA), radiofrequency ablation (RFA), and stereotactic body radiation therapy (SBRT). More than 85% of studies on these technologies are single arm, making head-to-head analysis methods challenging due to broken evidence networks. The objective of this research was to explore methods that can be used to pool and compare such large volumes of single arm data.

METHODS: A systematic literature review was performed to identify comparative and single‑arm studies assessing MWA, RFA, and SBRT in lung cancer patients. Overall survival and local tumor progression were assessed at several time-points. Several methodological techniques were compared, including single-arm random effects meta-analysis, univariable and multivariable meta-regressions, single-arm Bayesian meta-analysis, and propensity-score matched (PSM) analysis.

RESULTS: First, single-arm random effects meta-analyses provided pooled treatment point estimates to inform benchmarks; however, comparisons between outcomes could not be made due to the strong risk of confounding. Second, univariable and multivariable meta-regressions were conducted to better compare treatments, but data gaps limited the ability to control for some confounding variables. Third, single-arm Bayesian meta-analysis enabled treatment ranking with percentage likelihood one treatment is better than another; however, this method was sensitive to priors when the sample size was small. Fourth, PSM analysis involved the creation of simulated comparative trials, based on matched select covariates, enabling meta‑analyses through complete evidence networks. However, this method was limited by treatments with the fewest study arms.

CONCLUSIONS: The methods assessed enabled the indirect comparison of treatments that have been evaluated primarily by single-arm studies. Each method has strengths and limitations and by describing those here, we have provided a guide for other researchers to select the most appropriate approach for the meta‑analysis of single-arm data to allow comparisons between treatment arms.