Clinical study report data did not substantially alter point estimates but improved precision in a nephrology systematic review

Abstract

Objectives: To investigate the effect of adding clinical study report (CSR) data to publication and author data on data completeness and meta-analytical results. Study Design and Setting: Case report of a systematic review with meta-analysis of randomized controlled trials on icodextrin compared to glucose solutions in peritoneal dialysis including 19 clinical trials. We considered the outcomes mortality, peritoneal dialysis technique failure, quality of life, net peritoneal ultrafiltration (at 3e6 months, and 1e2 years), serious adverse events (SAE), peritonitis, and uncontrolled fluid overload. The results for these outcomes were reanalyzed using (a) publication and author data only, then compared with (b) publication and author data with added CSR data. At outcome level, we compared the number of included trials, pooled point estimates (ie, regarding effect direction), and 95% confidence intervals (CIs; ie, regarding overlap and width) between the two groups of trials (a and b). We illustrated the results of our meta-analyses in forest plots and narratively summarized them. Results: Except for two of the eight assessed outcomes (quality of life and net peritoneal ultrafiltration [1e2 years]), more complete data was available when adding CSRs to publication and author data. Point estimates were not statistically significantly different for publication and author data, compared to publication, author, and CSR data, for any outcome. For peritonitis, point estimates were on opposite sides of the line of no effect but remained statistically nonsignificant when adding CSR data. For SAE and net peritoneal ultrafiltration (3e6 months), the width of the 95% CI was narrower when adding CSR data and for net peritoneal ultrafiltration (3e6 months), in addition, the point estimate statistically significantly favored icodextrin when adding CSR data. Conclusion: The fraction of publications reporting results varied substantially by outcome, with SAE most under-reported in publications. While the integration of CSR data did not substantially alter meta-analytical results, it enhanced data completeness and precision in effect estimates. Our findings underscore the importance of accessing CSR data to optimize evidence syntheses and inform clinical decision- making.