Masclee GMC, Valkoff VE, Straatman H, Herings R, Garbe E, Schink T, Kollhorst B, Arfe A, Lucchi S, Villa M, Castellsague J, Perez-Gutthann S, Varas-Lorenzo C, Schade R, Schuemie M, Vergouwe Y, Steyerberg E, Sturkenboom M, Romio S. Individualized NSAID prescribing based on gastrointestinal and cardiovascular risks: a decision model in the SOS project. Poster presented at the 31st International Conference on Pharmacoepidemiology and Therapeutic Risk Management; August 2015. Boston, MA. [abstract] Pharmacoepidemiol Drug Saf. 2015 Sep; 24(S1):36-7. doi: 10.1002/pds.3838.


Background: Use of non-steroidal anti-inflammatory drugs (NSAIDs) can increase the risk of upper gastrointestinal complications (UGIC) and cardiovascular (CV) events. However, the risk may differ between individual NSAIDs and subjects. Decision models for selecting the safest NSAID to treat individual patients are not available.

Objectives: To provide a decision analytic model to guide physicians in the choice of NSAID treatment that would yield the lowest GI and CV risk for individual patients.

Methods:
Design: We produced a decision model integrating information from 1) a case-control study on individual NSAIDs and UGIC and CV events; 2) a risk function for patient characteristics associated with UGIC and CV events; 3) disutility weights at 4 weeks: 0.54 for UGIC, 0.65 for ischemic stroke (iCVA), 0.63 for acute myocardial infarction (AMI), and 0.29 for heart failure (HF).

Setting: Six European healthcare datasources: (IPCI, PHARMO (NL); SISR, OSSIFF (Italy); GePaRD (Germany) and THIN (UK).

Datasource specific study period: 1999-2011.

Exposure: Thirteen NSAIDs.

Outcome: UGIC, iCVA, AMI and HF hospitalization

Statistical analysis: Adjusted odds ratios (ORs) for UGIC and CV events during individual NSAID exposure. Poisson regression model for risk function. A decision tree was used for the decision model.


Results:
In the case-control study, 23,411 UGIC, 35,691 iCVA, 68,757 AMI, and 79,876 HF cases were identified among 8.9 million new NSAID users. The lowest risks were seen for use of celecoxib for UGIC (OR=1.1) and for HF (OR=1.0), for iCVA for ketoprofen (OR=0.9) and for AMI for tenoxicam and aceclofenac (OR=1.0). For all outcomes ketorolac yielded the highest risks. In the risk function, for each outcome, age was the most important predictor, followed by a prior history of the event and sex. In the final decision model, over different scenarios, most preferable NSAIDs were aceclofenac and celecoxib, thereafter nimesulide and ibuprofen. Piroxicam and ketorolac were the least preferable NSAIDs.

Conclusions: The SOS study provided an integrated GI and CV safety decision model for new NSAID users, which may guide physicians in clinical decision making.

Share on: