Table 5.

Deterministic sensitivity analyses undertaken

Sensitivity analysis	Change tested	Reason
Observational data	Use observational data for adverse events rather than randomised controlled trial data
Stroke risk	Alter the stroke risk estimates such that both COX 2 selective drugs (celecoxib and etoricoxib) have the same risk	High uncertainty in the evidence
Dose of NSAIDs	Assume diclofenac dose of 150 mg, rather than 100 mg Assume naproxen dose of 500 mg or 1000 mg rather than 750 mg Assume ibuprofen dose of 2400 mg rather than 1200 mg.
Heart failure risk	Assume same risk of heart failure for all treatment options	High uncertainty in the evidence
Myocardial infarction risk	Assume ibuprofen, naproxen, and diclofenac are all associated with the same risk of myocardial infarction	High uncertainty in the evidence
Hip fracture associated with proton pump inhibitor usage	Assume increased risk of hip fracture with proton pump inhibitor usage	New data51 52
Proton pump inhibitor risk reduction	Alter the relative risk reduction estimated to result from addition of a proton pump inhibitor to an NSAID or a COX 2 selective inhibitor
Dose-adverse event relation	Alter the assumption that a 50% reduction in dose leads to a 25% reduction in adverse events