To critique the methodological shortcomings in Gu et al's Bayesian network meta-analysis on telerehabilitation for chronic low back pain.
Approach:
Risk-of-Bias Assessment: Critique of the misapplication of the Cochrane Risk of Bias (RoB) 2.0 tool, which was not supported by the software used.
Statistical Heterogeneity: Discussion on the inappropriate use of I2 for assessing heterogeneity instead of τ in Bayesian NMA.
Baseline Imbalances: Highlighting significant baseline imbalances in age and severity among treatment groups, questioning the comparability of populations.
Convergence Evaluation: Critique of the lack of formal diagnostics for Markov chain Monte Carlo convergence, impacting the reliability of estimates.
Prediction Intervals: Concerns regarding the use of fixed τ values for prediction intervals, which deviates from Bayesian methodology.
Key Findings:
Misrepresentation of the risk-of-bias assessment tool used in the analysis.
Inappropriate application of I2 for heterogeneity assessment in Bayesian NMA.
Significant baseline imbalances in age and severity among treatment groups.
Lack of formal diagnostics for evaluating convergence in Bayesian analysis.
Interpretation:
The methodological flaws identified raise concerns about the validity and reliability of the findings from the meta-analysis.
Limitations:
Misapplication of statistical tools may mislead readers.
Absence of necessary diagnostics prevents assessment of the robustness of conclusions.
Pooling of populations without sensitivity analyses raises questions about the validity of indirect comparisons.
Conclusion:
The authors are encouraged to address these methodological concerns to enhance the reliability of their findings.
