Katie Carmichael1, Sarah J Nolan2, Jennifer Weston3, Catrin Tudur Smith4, Anthony G Marson5. 1. Department of Molecular and Clinical Pharmacology, Clinical Sciences Centre for Research and Education, University of Liverpool, Lower Lane, Liverpool L9 7LJ, United Kingdom. Electronic address: katie_carmichael@hotmail.com. 2. MRC North West Hub for Trials Methodology Research, Department of Biostatistics, University of Liverpool, 1st Floor Duncan Building, Daulby Street, Liverpool L69 3GA, United Kingdom. Electronic address: sjn16@liverpool.ac.uk. 3. Department of Molecular and Clinical Pharmacology, Clinical Sciences Centre for Research and Education, University of Liverpool, Lower Lane, Liverpool L9 7LJ, United Kingdom. Electronic address: jennifer.weston@liverpool.ac.uk. 4. MRC North West Hub for Trials Methodology Research, Department of Biostatistics, University of Liverpool, 1st Floor Duncan Building, Daulby Street, Liverpool L69 3GA, United Kingdom. Electronic address: cat1@liverpool.ac.uk. 5. Department of Molecular and Clinical Pharmacology, Clinical Sciences Centre for Research and Education, University of Liverpool, Lower Lane, Liverpool L9 7LJ, United Kingdom. Electronic address: marjon01@liverpool.ac.uk.
Abstract
PURPOSE: Treatment decisions should be informed by high quality evidence of both the potential benefit and harms of treatment alternatives. Randomised controlled trials (RCTs) provide the best evidence regarding benefits; however information relating to serious, rare and long-term harms is usually available only from non-randomised studies (NRSs). The aim of this study was to use a checklist based on the CONSORT (Consolidating Standards for Reporting Trials) extension for harms recommendations to assess the quality of reporting of harms data in both NRSs and RCTs of antiepileptic drugs, using studies of topiramate as an example. RESULTS: Seventy-eight studies were included from an online search of seven databases. Harms data was extracted from each study using a 25-point checklist. The mean number of items met was 11.5 (SD 2.96) per study. Commercially funded studies met on average 12.7 items and non-commercially funded studies met 10.08 (p value < 0.001). RCTs met on average 13.0 items and NRSs met 10.8 (p = 0.001). Multi-centre studies and commercially funded studies met significantly more items than single centre and non-commercially funded studies respectively. There was no significant difference in the mean number of items met by studies that had included adult vs. child participants, or studies published pre- vs. post-CONSORT extension for harms in 2004. CONCLUSIONS: Reporting of harms is significantly better in RCTs than in NRSs of TPM, but is suboptimal overall and has not improved since the publication of CONSORT extension for harms in 2004. There is a need to improve the reporting of harms in order to better inform treatment decisions.
PURPOSE: Treatment decisions should be informed by high quality evidence of both the potential benefit and harms of treatment alternatives. Randomised controlled trials (RCTs) provide the best evidence regarding benefits; however information relating to serious, rare and long-term harms is usually available only from non-randomised studies (NRSs). The aim of this study was to use a checklist based on the CONSORT (Consolidating Standards for Reporting Trials) extension for harms recommendations to assess the quality of reporting of harms data in both NRSs and RCTs of antiepileptic drugs, using studies of topiramate as an example. RESULTS: Seventy-eight studies were included from an online search of seven databases. Harms data was extracted from each study using a 25-point checklist. The mean number of items met was 11.5 (SD 2.96) per study. Commercially funded studies met on average 12.7 items and non-commercially funded studies met 10.08 (p value < 0.001). RCTs met on average 13.0 items and NRSs met 10.8 (p = 0.001). Multi-centre studies and commercially funded studies met significantly more items than single centre and non-commercially funded studies respectively. There was no significant difference in the mean number of items met by studies that had included adult vs. child participants, or studies published pre- vs. post-CONSORT extension for harms in 2004. CONCLUSIONS: Reporting of harms is significantly better in RCTs than in NRSs of TPM, but is suboptimal overall and has not improved since the publication of CONSORT extension for harms in 2004. There is a need to improve the reporting of harms in order to better inform treatment decisions.