The consequences of a wholly sponsor-generated evidence base
What is known about the direction and magnitude of the difference between sponsor-conducted and independently conducted trials, and how should an evidence body treat a class whose entire evidence base is sponsor-generated?
§1Abstract
§1.1Review question
What is known about the direction and magnitude of the difference between sponsor-conducted and independently conducted trials, and how should an evidence body treat a class whose entire evidence base is sponsor-generated?
§1.2PICO frame
Table 1. The registered PICO frame. Any change made after protocol registration is recorded in the amendment log at §5 and is identified there as post hoc.
| Element | As registered |
|---|---|
| Population | Randomised trials of pharmacological interventions. |
| Intervention | Sponsor funding and sponsor conduct. |
| Comparator | Independent funding or independent conduct. |
| Outcomes | Effect estimate; probability of a favourable conclusion; completeness of harms reporting. |
§1.3Method in brief
A review whose unit of analysis is a study characteristic rather than a treatment effect. The protocol was registered with the Institute's secretariat before the search was executed. The search was run on 13 March 2024 across 8 sources and is reproduced in full at §2. Screening, extraction and certainty assessment were performed independently by two members of the secretariat with disagreement resolved by a third.
§1.4Conclusion
Moderate certainty methodological evidence indicates that industry-sponsored trials report more favourable efficacy conclusions and less complete harms data than independently conducted trials of the same interventions, and that the difference is not fully explained by risk-of-bias domains as conventionally assessed. Every pivotal trial in the incretin class is sponsor-conducted. The Institute does not downgrade certainty for sponsorship as a separate domain, because doing so would double-count risk of bias; it records sponsorship on the front matter of every trial abstract instead, and treats the uniformity of sponsorship across the class as a limitation of the evidence base rather than of any individual trial.
The conclusion rests on 24 contributing studies, listed with their extracted data at §3 and summarised outcome by outcome at §4.
§1.5Limitations
§1.6Consultation
This review was released for public comment before ratification. Draft synthesis: The consequences of a wholly sponsor-generated evidence base received 10 submissions. Amendments arising are recorded in the amendment log and are traceable to a numbered submission.