Introduction
Methods
A comprehensive and systematic search of the literature was conducted in accordance with the PRISMA statement and PRISMA extension statement for NMAs. The search strategy and the study selection criteria were clearly stated. Data extraction and study quality assessment was done independently by two reviewers; inconsistencies were resolved by a third reviewer. The Cochrane collaboration risk of bias tool and the Newcastle-Ottawa scale were used to appraise the methodological quality of RCTs and non-RCTs, respectively. The quality of the included studies varied, with low, medium, and high risk of bias found depending on the trial. The outcome measures were selected appropriately and clearly described. The choice of fixed or random effect was justified and based on I2 value of <50% as a cutpoint for statistical heterogeneity. Both direct pairwise and indirect comparisons were performed to assess the consistency assumption, as well as the results from consistency and inconsistency models of NMA. Assessment of publication bias was done visually using Funnel plots and quantitatively using the Begg’s test and Egger’s test. Other than the pooled PFS and discontinuation rate, publication bias was unlikely for other outcomes. The methods for indirect comparison (Bayesian network meta-analyses) and ranking treatment probabilities for each line of therapy were appropriate. The methods for pooling Kaplan-Meier curves and other binary outcome measures were appropriate.
Results
Identification and selection of full-text studies for the NMA were well reported, as well as presented in a PRISMA flowchart. Additionally, a network diagram was provided. Although not in great detail, a table with study and patient characteristics was provided; and summary effect estimates from each included trial was available. Convergence of all models using the MCMC method was shown with a PSRF value of 1.00.
Discussion
A description of the main findings was presented that highlighted the potential limitations of the results as well as possible explanations for discrepancies across studies. The authors did not provide a discussion on the generalizability of findings; however, given the included studies combined a high number of patients, all available ALK-inhibitors, and trials conducted across the world in various settings, the generalizability of the results is not likely to be a concern.
| Methods
The models were conducted without covariate adjustment for patient or study characteristics, and hence control of potential bias could not be assessed. There was no information on prior distributions for model parameters, and whether priors were informative or non-informative. No sensitivity analyses were performed to assess the effect of different covariate distributions or model assumptions.
Discussion:
Between-study heterogeneity was high for a number of pairwise comparisons, due primarily to differences in doses of treatment, baseline parameters of enrolled patients, and follow-up duration. Overall, the original trials used to pool results for direct and indirect comparisons included a mix of ALK inhibitor-naive and pretreated patients, therefore the combined results should be interpreted with caution. However, this was not a limitation of the NMA per se, rather limited the usability of the results given the research questions in this Rapid Response report.
|
