To review the clinical effectiveness and cost-effectiveness of cyclooxygenase-2 (COX-2) selective non-steroidal anti-inflammatory drugs (NSAIDs) (etodolac, meloxicam, celecoxib, rofecoxib, etoricoxib, valdecoxib and lumiracoxib) for osteoarthritis (OA) and rheumatoid arthritis (RA). Electronic databases were searched up to November 2003. Industry submissions to the National Institute for Health and Clinical Excellence (NICE) in 2003 were also reviewed. Systematic reviews of randomised controlled trials (RCTs) and a model-based economic evaluation were undertaken. Meta-analyses were undertaken for each COX-2 selective NSAID compared with placebo and non-selective NSAIDs. The model was designed to run in two forms: the 'full Assessment Group Model (AGM)', which includes an initial drug switching cycle, and the 'simpler AGM', where there is no initial cycle and no opportunity for the patient to switch NSAID. Compared with non-selective NSAIDs, the COX-2 selective NSAIDs were found to be equally as efficacious as the non-selective NSAIDs (although meloxicam was found to be of inferior or equivalent efficacy) and also to be associated with significantly fewer clinical upper gastrointestinal (UGI) events (although relatively small numbers of clinical gastrointestinal (GI) and myocardial infarction (MI) events were reported across trials). Subgroup analyses of clinical and complicated UGI events and MI events in relation to aspirin use, steroid use, prior GI history and Helicobacter pylori status were based on relatively small numbers and were inconclusive. In the RCTs that included direct COX-2 comparisons, the drugs were equally tolerated and of equal efficacy. Trials were of insufficient size and duration to allow comparison of risk of clinical UGI events, complicated UGI events and MIs. One RCT compared COX-2 (celecoxib) with a non-selective NSAID combined with a gastroprotective agent (diclofenac combined with omeprazole); this included arthritis patients who had recently suffered a GI haemorrhage. Although no significant difference in clinical GI events was reported, the number of events was small and more such studies, where patients genuinely need NSAIDs, are required to confirm these data. A second trial showed that rofecoxib was associated with fewer diarrhoea events than a combination of diclofenac and misoprostol (Arthrotec). Previously published cost-effectiveness analyses indicated a wide of range of possible incremental cost per quality-adjusted life-year (QALY) gained estimates. Using the simpler AGM, with ibuprofen or diclofenac alone as the comparator, all of the COX-2 products are associated with higher costs (i.e. positive incremental costs) and small increases in effectiveness (i.e. positive incremental effectiveness), measured in terms of QALYs. The magnitude of the incremental costs and the incremental effects, and therefore the incremental cost-effectiveness ratios, vary considerably across all COX-2 selective NSAIDs. The base-case incremental cost per QALY results for COX-2 selective NSAIDs compared with diclofenac for the simpler model are: celecoxib (low dose) 68,400 pounds; celecoxib (high dose) 151,000 pounds; etodolac (branded) 42,400 pounds; etodolac (generic) 17,700 pounds; etoricoxib 31,300 pounds; lumiracoxib 70,400 pounds; meloxicam (low dose) 10,300 pounds; meloxicam (high dose) 17,800 pounds; rofecoxib 97,400 pounds; and valdecoxib 35,500 pounds. When the simpler AGM was run using ibuprofen or diclofenac combined with proton pump inhibitor (PPI) as the comparator, the results change substantially, with the COX-2 selective NSAIDs looking generally unattractive from a cost-effectiveness point of view (COX-2 selective NSAIDs were dominated by ibuprofen or diclofenac combined with PPI in most cases). This applies both to 'standard' and 'high-risk' arthritis patients defined in terms of previous GI ulcers. The full AGM produced results broadly in line with the simpler model. The COX-2 selective NSAIDs examined were found to be similar to non-selective NSAIDs for the symptomatic relief of RA and OA and to provide superior GI tolerability (the majority of evidence is in patients with OA). Although COX-2 selective NSAIDs offer protection against serious GI events, the amount of evidence for this protective effect varied considerably across individual drugs. The volume of trial evidence with regard to cardiovascular safety also varied substantially between COX-2 selective NSAIDs. Increased risk of MI compared to non-selective NSAIDs was observed among those drugs with greater volume of evidence in terms of exposure in patient-years. Economic modelling shows a wide range of possible costs per QALY gained in patients with OA and RA. Costs per QALY also varied if individual drugs were used in 'standard' or 'high'-risk patients, the choice of non-selective NSAID comparator and whether that NSAID was combined with a PPI. With reduced costs of PPIs, future primary research needs to compare the effectiveness and cost-effectiveness of COX-2 selective NSAIDs relative to non-selective NSAIDs with a PPI. Direct comparisons of different COX-2 selective NSAIDs, using equivalent doses, that compare GI and MI risk are needed. Pragmatic studies that include a wider range of people, including the older age groups with a greater burden of arthritis, are also necessary to inform clinical practice.
