Evaluation of risk profiles for gastrointestinal and cardiovascular adverse effects in nonselective NSAID and COX-2 inhibitor users - A cohort study using pharmacy dispensing data in the Netherlands

Background: Newly approved drugs, in comparison with older drugs, are more often prescribed to patients who have not responded satisfactorily to established related drugs or as first-line therapy to patients with a high baseline risk for adverse outcomes (i.e. channelling). However, these patients are less likely to benefit from the prescribed drug and/or are more prone to adverse drug reactions. Therefore, it is difficult to unravel whether observed risks or increases in risk of new drugs are real, i.e. related to the pharmacology, or whether these are related to selective prescribing to patients who are more susceptible to adverse events because of some underlying risk factor(s). The channelling paradox may exist for cyclo-oxygenase (COX)-2 selective inhibitors ('coxibs') instead of traditional nonselective NSAIDs in relation to both gastrointestinal (GI) and cardiovascular (CV) safety. Objective: To evaluate the risk profiles for GI and CV adverse effects in nonselective NSAID and coxib new-user populations over time, in terms of a quantitative measure since the introduction of coxibs. Methods: This was a population-based cohort study using the Dutch pharmaceutical claims database (Foundation for Pharmaceutical Statistics). Eligible patients (>= 18 years) were those where the date of their first prescription (index date) of an NSAID (first-line [e.g. ibuprofen] or second-line [e.g. piroxicam] nonselective NSAID, COX-2 preferential NSAID or coxib) was between January 1999 and December 2003. For each patient, GI and CV risk profiles at index date were defined by a cumulative score derived from dispensing data (patient age, sex and history of medication use within 6 months of index date). Risk scores were categorized as low (score = 0), medium (1) or high (2+). Patients were recorded as switchers based on other NSAID use prior to the index date. Other information collected included the Chronic Disease Score (CDS). Crude odds ratios (ORs) were calculated for risk factors for each NSAID group versus first-line nonselective NSAID users as the reference cohort. The effect of calendar time was examined by plotting mean CV or GI risk score by quarter-year. Correlation between GI and CV scores was examined using the Pearson correlation coefficient (R). Data were stratified by patients' history of switching. Results: The four cohorts comprised patients using: first-line nonselective NSAIDs (n = 42 750); second-line nonselective NSAIDs (n = 1771); COX-2 preferential NSAIDs (n = 3661) and coxibs (n = 4861) patients. New coxib users were most likely to have high GI and CV risk scores (OR 5.3 [95% CI 5.0, 5.6] and OR 2.2 [95% CI 2.1, 2.4], respectively). At the individual patient level, GI and CV risk profiles were moderately well correlated for all NSAID cohorts (R range 0.48 to 0.62). There was no remarkable change in mean GI or CV risk profile of patients over calendar time since the market introduction of coxibs. Discussion: Of the four NSAID cohorts, new coxib users tended to have the highest numbers of GI and CV risk factors, with no obvious change over calendar time. There was also evidence of correlation between GI and CV risk scores. Thus, selective prescribing of coxibs applies to people with co-existing CV as well as GI risk factors. This is important when comparing the safety and/or efficacy of new therapies to existing therapies, and emphasizes the difficulties encountered by prescribers in assessing levels of risk when initiating coxib treatment.