Objective:

To analyze the influence of estimation selection on the observed mortality threshold associated with very-low-dose continuous renal replacement therapy (CRRT).

Approach:

• Analysis Reconstruction: Reconstructed the analysis of Lumlertgul et al. to assess how different estimates affect the observed mortality threshold.
• Comparison of Models: Contrasted conventional Cox models with marginal structural Cox models to evaluate the impact of treatment and confounding factors.
• Sensitivity Analysis: Conducted sensitivity analyses to explore the effects of different estimates on pooled results.

Key Findings:

• The apparent mortality threshold of 13 mL/kg/h varies based on the selected estimate according to the analysis.
• Conventional analysis yielded a hazard ratio (HR) of 1.70, while marginal structural analysis indicated an HR of 1.08.
• Neither time-dependent estimate reached statistical significance, suggesting adjustment for time-dependent factors nullified the baseline association.
• The review's ROBINS-I assessment indicated serious risk of bias from selection of reported results.

Interpretation:

The findings suggest that the choice of estimation method significantly influences the observed association between CRRT dose and mortality.

Limitations:

• The review protocol lacked clear guidelines for selecting among multiple adjusted estimates.
• Significant heterogeneity in meta-analysis results exceeded the threshold for reporting.

Conclusion:

The analysis highlights the need for clarity in estimate selection and reporting.

Sources:

• Lumlertgul et al. Study
• Faria and Libório Analysis
• Okamoto et al. Study
• PRISMA Guidelines
• Methodological Guidance