Article type
Year
Abstract
Background: Surgical trials pose special challenges in design, analysis, and reporting. The number of surgical procedures used in neurosurgical practice is already large and rapidly increasing both for traditional and new techniques. Neurosurgical procedures need to be evaluated with randomized controlled trials (RCTs) that are carefully performed and reported.
Objectives: To appraise systematically the study design and quality of reporting of randomized controlled trials (RCTs) on neurosurgical procedures and to identify potential defects and biases.
Methods: RCTs with at least 5 patients comparing any neurosurgical procedure against another procedure, non-surgical treatment or no treatment were retrieved from MEDLINE, EMBASE, and the Cochrane Library. Analysis was limited to trials involving the brain or spinal cord. The pertinence of surgical procedures to neurosurgical practice was judged by two expert neurosurgeons who evaluated whether a specific procedure was likely to be performed by neurosurgeons even if other surgical specialists may perform it as well. Two investigators independently extracted data. Differences were solved by consensus and a third investigator settled any remaining discrepancies. We analysed study design, quality of reporting, and trial results.
Results: Median sample size in the 108 eligible reports was 68 patients. Ninety-nine trials (91.7%) reported inclusion and exclusion criteria, 55 (50.9%) mentioned the randomisation mode, and 87 (80.6%) described adequately withdrawals, but only 31 (28.7%) described allocation concealment, only 23 (21.3%) gave power calculations, and only 20 (18.5%) were adequately powered. Significant efficacy or trend for efficacy was claimed in 46 reports (42.6%) while no difference between the compared procedures was found in 60 trials (55.6%). Trials with larger sample size were more likely to report withdrawals (p=0.02) and power calculations (p=0.006). Only 14 trials (13.6%) were double blind and this was less frequent in trials with longer follow-up (p=0.02). Among quality criteria, only the reporting of randomisation mode improved significantly over time (p = 0.015).
Conclusions: Several aspects of the design and reporting of RCTs on neurosurgical procedures can be improved. Trials with longer follow-up, adequately powered, and accurately reported are needed.
Objectives: To appraise systematically the study design and quality of reporting of randomized controlled trials (RCTs) on neurosurgical procedures and to identify potential defects and biases.
Methods: RCTs with at least 5 patients comparing any neurosurgical procedure against another procedure, non-surgical treatment or no treatment were retrieved from MEDLINE, EMBASE, and the Cochrane Library. Analysis was limited to trials involving the brain or spinal cord. The pertinence of surgical procedures to neurosurgical practice was judged by two expert neurosurgeons who evaluated whether a specific procedure was likely to be performed by neurosurgeons even if other surgical specialists may perform it as well. Two investigators independently extracted data. Differences were solved by consensus and a third investigator settled any remaining discrepancies. We analysed study design, quality of reporting, and trial results.
Results: Median sample size in the 108 eligible reports was 68 patients. Ninety-nine trials (91.7%) reported inclusion and exclusion criteria, 55 (50.9%) mentioned the randomisation mode, and 87 (80.6%) described adequately withdrawals, but only 31 (28.7%) described allocation concealment, only 23 (21.3%) gave power calculations, and only 20 (18.5%) were adequately powered. Significant efficacy or trend for efficacy was claimed in 46 reports (42.6%) while no difference between the compared procedures was found in 60 trials (55.6%). Trials with larger sample size were more likely to report withdrawals (p=0.02) and power calculations (p=0.006). Only 14 trials (13.6%) were double blind and this was less frequent in trials with longer follow-up (p=0.02). Among quality criteria, only the reporting of randomisation mode improved significantly over time (p = 0.015).
Conclusions: Several aspects of the design and reporting of RCTs on neurosurgical procedures can be improved. Trials with longer follow-up, adequately powered, and accurately reported are needed.