Points to consider for designing, conducting and interpreting comparative effectiveness trials in rheumatology
| 1. Trial design (protocol) | Register every trial protocol in detail in http://www.clinicaltrials.gov or similar registries |
| Carefully weigh the pros and cons of a randomised trial versus an observational trial with statistical adjustment and modelling | |
| Think about the most appropriate design template (superiority vs Non-inferiority), motivate it and describe the design If non-inferiority: discuss all aspects of the non-inferiority margin If superiority: discuss all aspects of the minimum clinically important difference | |
| Realise all implications of non-blinding or assessor (single) blinding versus assessor and patient (double) blinding before making a decision about an open trial, a partially or a fully blinded trial | |
| Describe content and timing of the primary end point(s) | |
| Describe a sound statistical analysis plan for the primary end point, including unequivocal wording about intention-to-treat vs completers analysis and about sensitivity analysis | |
| Provide a sound and realistic sample size calculation including implications for statistical power based on the predefined minimum clinically important difference or non-inferiority margin | |
| 2. Trial conduct | Adhere to the trial protocol registered in http://www.clinicaltrials.gov or other registries. Do not amend key criteria of the trial protocol |
| Do not stop the inclusion of the trial before the planned inclusion is finished | |
| Do all you can to obtain outcome data from all patients, either in the trial or withdrawn | |
| Do everything to convince your clinical investigators to adhere to the agreed treatment protocol and to document all their actions and violations appropriately | |
| 3. Trial report | Analyse the data entirely in accordance with the statistical analysis plan and report accordingly. Make clear distinctions between preplanned and unplanned (posthoc) analyses |
| Resist the desire to report interim analyses because of ‘promising data’ | |
| Do not draw ‘non-inferiority’ or ‘equivalence’ conclusions from trials with a ‘superiority’ trial design | |
| Do provide ‘superiority conclusions’ (as a secondary conclusion) if ‘non-inferiority trials’ give evidence of this | |
| Make clear in the discussion section where the reported trial has violated the ‘ideal protocol’ and what the repercussions are for internal and external validity (‘generalisability’) | |
| Refrain from firm conclusions that are not backed by robust results. Inclusion of other considerations, such as pharmacoeconomics, are relevant to the discussion of the ultimate use of the data, but they should not be used to imply validity of the data | |
| Refrain from firm conclusions that are tendentious in the title, abstract and discussion section | |
| Always make clear that fallible trial methodology is among the explanations for a certain result. Echo the opinion that ‘comparative clinical research’ may be upscaled observational research |