To evaluate the impact of rofecoxib withdrawal on the characteristics of patients prescribed selective cyclooxygenase 2 (COX-2) inhibitors.
To evaluate the impact of rofecoxib withdrawal on the characteristics of patients prescribed selective cyclooxygenase 2 (COX-2) inhibitors.
The General Practice Research Database was used to identify patients age ≥18 years who were prescribed a selective COX-2 inhibitor. Various patient characteristics were noted at the start of therapy: age, sex, nonsteroidal antiinflammatory drug–related risk factors for upper gastrointestinal (GI) events, and the Framingham risk score for cardiovascular disease. Logistic regression was used to compare patients using selective COX-2 inhibitors before and after September 2004.
The study population included 171,645 patients receiving selective COX-2 inhibitors. The number of users substantially increased over time until September 2004 and sharply declined thereafter. Approximately 80% stopped selective COX-2 inhibitor therapy within 6 months. Patients receiving selective COX-2 inhibitors after September 2004 were younger and included more men compared with those receiving therapy before September 2004. There was no change before and after September 2004 in the proportion of patients with GI risk factors or high Framingham risk scores, after adjustment for age and sex. A correlation was found between presence of GI risk factors and high Framingham risk scores. Only 20% of patients receiving selective COX-2 inhibitors had GI risk factors but low Framingham risk score, which did not change after September 2004.
There was no channeling in the usage of selective COX-2 inhibitors toward patients with a high risk of GI and low risk of cardiovascular disease following the withdrawal of rofecoxib.
Before September 2004, selective cyclooxygenase 2 (COX-2) inhibitors ranked among the most commonly used medications in the world, with more than 1 million patients prescribed this type of nonsteroidal antiinflammatory drug (NSAID) in the UK alone (including rofecoxib, celecoxib, etoricoxib, and valdecoxib). Compared with traditional NSAIDs, selective COX-2 inhibitors were thought to have lower risks of upper gastrointestinal (GI) toxicity (1, 2). However, following the finding of an increased risk of myocardial infarction (MI), the selective COX-2 inhibitor rofecoxib was withdrawn from medical practice in September 2004 (1). Subsequently, the European Medicines Agency contraindicated the use of other selective COX-2 inhibitors in patients with ischemic heart disease or stroke and the use of etoricoxib in patients with uncontrolled hypertension. A warning was introduced for health professionals to exercise caution when prescribing selective COX-2 inhibitors for patients with risk factors for heart disease, such as hypertension, hyperlipidemia (high cholesterol levels), diabetes, and smoking, as well as for patients with peripheral arterial disease. Health professionals were also advised to prescribe the lowest effective dose of selective COX-2 inhibitors for the shortest possible duration (3).
The objective of the present study was to evaluate the trends in the characteristics of UK patients starting therapy with selective COX-2 inhibitors before and after the 2004 withdrawal of rofecoxib. Although there have since been studies describing patients receiving selective COX-2 inhibitors, none have evaluated any changes following the withdrawal of rofecoxib (4–6).
Information for the study was obtained from the General Practice Research Database (GPRD), which comprises the computerized medical records of general practitioners (GPs) in the UK. GPs play a key role in the UK health care system, as they are responsible for primary health care and specialist referrals. Patients are semipermanently affiliated with a practice, which centralizes the medical information from the GPs, specialist referrals, and hospitalizations. The data recorded in the GPRD include demographic information, prescription details, clinical events, preventive care provided, specialist referrals, hospital admissions, and patients' major outcomes (7). In the UK, GPs are routinely informed about specialist care and hospitalizations. Several independent validation studies have confirmed a high level of completeness and validity in the GPRD, specifically with regard to MI and upper GI events (8, 9). Data collection for the GPRD started in 1987 and ended, for the present study, in January 2006.
The study population comprised men and women in the GPRD age ≥18 years who started therapy with selective COX-2 inhibitors (i.e., first-ever prescription for selective COX-2 inhibitors). Various patient characteristics were assessed at the start of selective COX-2 inhibitor therapy, including age; sex; medical history of ischemic heart disease; and prescription of antihypertensives, antiplatelets or anticoagulants, and aspirin in the previous 6 months. The number of major risk factors for NSAID-related upper GI events and the Framingham risk score for cardiovascular disease were also determined for each patient at the start date. The upper GI risk factors included age ≥65 years; recent prescription of oral glucocorticoids, anticoagulants, or acetylsalicylic acid in the previous 6 months; and a history of peptic upper GI bleeding (4). The Framingham risk score was based on the weighing and scoring of the following risk factors: sex, age, systolic blood pressure, smoking, total cholesterol divided by high-density lipoprotein cholesterol, diabetes mellitus, and left ventricular hypertrophy (10). The population mean was used for patients without blood pressure and cholesterol values. In accordance with the Framingham equations, we classified former smokers who had stopped smoking within the previous 4 years as current smokers and former smokers who had stopped smoking more than 4 years earlier as nonsmokers. Each patient was stratified according to calendar time (6 monthly intervals) at the start date of therapy with selective COX-2 inhibitors.
The incidence rates of MI or unstable angina and upper GI events (gastroduodenal ulcers and complications such as upper GI hemorrhage) were estimated in this population. Patients were observed for the occurrence of these events from the start of therapy with selective COX-2 inhibitors until 3 months after the expected end of use, occurrence of the outcome, or the date of censoring (i.e., end of the patient's followup), whichever date came first. The incidence rate was estimated by dividing the number of cases by the number of person-years of followup. A 5-year cumulative incidence was estimated by converting this annualized rate into a cumulative incidence over 5 years.
An analysis of therapy duration (i.e., persistence) was conducted by evaluating repeat prescribing over time. If a patient did not receive a repeat prescription within 3 months after the expected end of the treatment course, the patient was considered to have discontinued treatment with selective COX-2 inhibitors at the expected end of treatment. Life-table analysis was used. A sensitivity analysis was conducted using a shorter window of 1 month instead of 3 months for repeat prescribing.
Logistic regression was used to compare patients who started selective COX-2 therapy in the 6 months preceding the withdrawal of rofecoxib (January 1, 2004 to June 31, 2004) with patients who started after the withdrawal of rofecoxib (July 1, 2005 to December 31, 2005). Odds ratios (ORs) and 95% confidence intervals were estimated.
The study population included 171,645 patients receiving selective COX-2 inhibitors who were age ≥18 years. Figure 1 shows that the use of selective COX-2 inhibitors substantially increased over time until September 2004. After this date, use sharply decreased. During January to June 2004, 29% of all NSAID prescriptions were for selective COX-2 inhibitors, and following the withdrawal of rofecoxib, their use fell to 7% (July to December 2005).
The baseline characteristics of the patients are shown in Table 1. A total of 6.2% of the patients had a history of peptic ulcer or upper GI bleeding. Of these patients, 88.9% had the event at least 2 years prior to the start of selective COX-2 inhibitors. A total of 47.0% of the patients were started on rofecoxib and 42.5% started celecoxib. The move toward prescribing selective COX-2 inhibitors to younger patients represented the largest change in prescribing patterns after the withdrawal of rofecoxib. Although proportionally, fewer patients with major risk factors for GI and MI events received selective COX-2 inhibitors after the withdrawal of rofecoxib, this change could be mostly attributed to the change in prescribing to younger patients. The reductions in the use of selective COX-2 inhibitors were proportionally largest in Scotland and smallest in Northern Ireland.
|Total sample (n = 171,645)||Comparison of patients before and after withdrawal of rofecoxib|
|January 2004 to June 2004||July 2005 to December 2005||Age- and sex-adjusted OR (95% CI)|
|18–39||18,004 (10.5)||2,894 (11.6)||421 (18.6)||Reference|
|40–64||74,814 (43.6)||10,894 (43.6)||1,152 (50.9)||0.73 (0.65–0.82)|
|≥65||78,827 (45.9)||11,192 (44.8)||689 (30.5)||0.43 (0.38–0.49)|
|Female||109,912 (64.0)||15,633 (62.6)||1,284 (56.8)||Reference|
|Male||61,733 (36.0)||9,347 (37.4)||978 (43.2)||1.24 (1.13–1.35)|
|London/South England||61,169 (35.6)||9,305 (37.2)||738 (32.6)||Reference|
|Middle England/Wales||54,051 (31.5)||7,730 (30.9)||802 (35.5)||1.29 (1.16–1.43)|
|North England||36,347 (21.2)||5,192 (20.8)||455 (20.1)||1.05 (0.93–1.19)|
|Scotland||10,432 (6.1)||1,347 (5.4)||68 (3.0)||0.58 (0.45–0.75)|
|Northern Ireland||9,646 (5.6)||1,406 (5.6)||199 (8.8)||1.58 (1.34–1.87)|
|Peptic ulcer/upper GI bleeding|
|No||161,077 (93.8)||23,699 (94.9)||2,155 (95.3)||Reference|
|Yes||10,568 (6.2)||1,281 (5.1)||107 (4.7)||0.99 (0.81–1.22)|
|Ischemic heart disease|
|No||149,898 (87.3)||22,017 (88.1)||2,069 (91.5)||Reference|
|Yes||21,747 (12.7)||2,963 (11.9)||193 (8.5)||0.87 (0.75–1.03)|
|No||106,894 (62.3)||15,518 (62.1)||1,571 (69.5)||Reference|
|Yes||64,751 (37.7)||9,462 (37.9)||691 (30.5)||0.97 (0.87–1.07)|
|Antiplatelets or anticoagulants|
|No||139,403 (81.2)||20,029 (80.2)||1,944 (85.9)||Reference|
|Yes||32,242 (18.8)||4,951 (19.8)||318 (14.1)||0.97 (0.87–1.07)|
|No||148,607 (86.6)||21,531 (86.2)||2,036 (90.0)||Reference|
|Yes||23,038 (13.4)||3,449 (13.8)||226 (10.0)||0.90 (0.78–1.05)|
As shown in Table 2, patients with ≥1 GI risk factor had a higher incidence of GI events and a higher risk of MI/unstable angina. Similarly, patients with higher Framingham risk scores for cardiovascular disease had increased rates of MI/unstable angina and upper GI events. A small percentage of patients receiving selective COX-2 inhibitors (19.7%) had ≥1 risk factor for GI events and low Framingham risk scores. However, the incidence of MI/unstable angina was also higher in this group due to the presence of other risk factors for MI.
|No. (%)||Age ≥65 years||Women||6-month persistence||5-year cumulative incidence of upper GI events||5-year cumulative incidence of MI/unstable angina|
|No GI risk factors||80,141 (46.7)||0.0||63.0||19.4||0.3||0.8|
|1 GI risk factor||62,252 (36.3)||81.0||67.3||26.4||3.3||3.5|
|≥2 GI risk factors||29,252 (17.0)||96.9||59.9||27.9||6.5||7.7|
|Very high||30,662 (17.9)||91.1||39.8||27.0||3.3||5.5|
|No GI risk factor and low/middle Framingham score||72,024 (42.0)||0.0||66.7||18.7||0.3||0.6|
|No GI risk factor and high/very high Framingham score||8,117 (4.7)||0.0||30.3||24.3||0.6||2.5|
|≥1 GI risk factor and low/middle Framingham score||33,757 (19.7)||71.0||81.7||25.9||5.6||3.8|
|≥1 GI risk factor and high/very high Framingham score||57,747 (33.6)||95.0||55.1||27.3||3.3||5.3|
The changes in the Framingham risk score and upper GI risk factors following rofecoxib withdrawal are shown in Table 3. After adjustment for age and sex, patients who started therapy with selective COX-2 inhibitors after rofecoxib withdrawal had a similar number of upper GI risk factors compared with those who started therapy before withdrawal. The proportion of patients with a very high Framingham risk score was similar before and after rofecoxib withdrawal (age- and sex-adjusted OR 1.14). There were no differences in age- and sex-adjusted OR between patients with and without missing values in the Framingham risk score for smoking, blood pressure, and cholesterol (OR 1.07 and 1.15, respectively).
|January 2004 to June 2004 (n = 24,980)||July 2005 to December 2005 (n = 2,262)||Overall||Indication||History of use of nonselective NSAIDs|
|No GI risk factors||14,072 (56.3)||1,494 (66.0)||Reference||Reference||Reference||Reference||Reference|
|1 GI risk factor||8,133 (32.6)||577 (25.5)||0.91 (0.81–1.02)||0.92 (0.75–1.12)||0.91 (0.79–1.04)||0.96 (0.77–1.20)||0.90 (0.78–1.02)|
|≥2 GI risk factors||2,775 (11.1)||191 (8.4)||0.92 (0.77–1.09)||0.83 (0.63–1.10)||0.99 (0.80–1.23)||1.20 (0.88–1.63)||0.81 (0.66–1.00)|
|Low||9,003 (36.0)||1,061 (46.9)||Reference||Reference||Reference||Reference||Reference|
|Middle||6,760 (27.1)||562 (24.8)||1.01 (0.89–1.16)||1.09 (0.83–1.44)||0.99 (0.85–1.16)||0.94 (0.72–1.23)||1.06 (0.91–1.24)|
|High||4,876 (19.5)||311 (13.7)||0.90 (0.75–1.08)||1.07 (0.76–1.50)||0.84 (0.68–1.05)||0.81 (0.57–1.16)||0.97 (0.79–1.20)|
|Very high||4,341 (17.4)||328 (14.5)||1.14 (0.93–1.39)||1.30 (0.88–1.90)||1.11 (0.88–1.41)||1.19 (0.80–1.76)||1.16 (0.91–1.48)|
|No GI risk factor and low/middle Framingham score||11,153 (44.6)||1,276 (56.4)||Reference||Reference||Reference||Reference||Reference|
|No GI risk factor and high/very high Framingham score||2,919 (11.7)||218 (9.6)||1.00 (0.84–1.20)||1.35 (0.95–1.91)||0.90 (0.72–1.2)||1.04 (0.76–1.44)||1.04 (0.83–1.30)|
|≥1 GI risk factor and low/middle Framingham score||4,610 (18.5)||347 (15.3)||0.92 (0.80–1.06)||1.05 (0.81–1.36)||0.88 (0.74–1.05)||1.06 (0.79–1.43)||0.89 (0.75–1.04)|
|≥1 GI risk factor and high/very high Framingham score||6,298 (25.2)||421 (18.6)||0.90 (0.77–1.06)||1.01 (0.75–1.36)||0.90 (0.74–1.09)||1.02 (0.75–1.38)||0.88 (0.73–1.06)|
Figure 2 illustrates the change in characteristics of patients receiving selective COX-2 inhibitors over calendar time and shows that, after withdrawal of rofecoxib, there was an increase in the percentage of younger patients and men receiving selective COX-2 inhibitors. The duration of treatment with selective COX-2 inhibitors was generally short and decreased after rofecoxib withdrawal. Using a 3-month time window for repeat prescribing, 77.2% of the patients stopped within 6 months; with a 1-month window, the proportion was 87.6%.
The overall changes in the incidence rates of MI and upper GI events are shown in Figure 3. The incidence rates of MI and upper GI events decreased over calendar time both before and after the withdrawal of rofecoxib.
This study demonstrated that the characteristics of patients and the extent of use of selective COX-2 inhibitors changed considerably after withdrawal of rofecoxib and the related publicity and regulatory action. Consequently, selective COX-2 inhibitors are now used more for younger patients and for shorter periods of time. The incidence of upper GI events in the population using selective COX-2 inhibitors decreased after the withdrawal of rofecoxib.
The withdrawal of rofecoxib from the market was unexpected and generated considerable publicity in both lay and medical press. Regulatory authorities also issued guidelines to restrict the use of the remaining selective COX-2 inhibitors. Our findings suggest that as a consequence, the selective COX-2 inhibitors are being used for shorter periods, consistent with the guidance from regulatory authorities, and are used more by younger patients at lower risk of MI. However, we could find no evidence after September 2004 to suggest the channeling (11) of selective COX-2 inhibitors to patients with high GI risk but low MI risk, maximizing the benefits and minimizing any cardiovascular risks of treatment. In fact, the cumulative incidence of both MI and upper GI events decreased in patients treated with selective COX-2 inhibitors after the withdrawal of rofecoxib. Based on our data, it appears that the GI and MI risks were strongly correlated in patients receiving selective COX-2 inhibitors, and therefore it may be difficult for prescribers to identify those patients with an increased risk of NSAID-related GI complications but good cardiovascular health. This may indicate that the current approach to minimization of MI risk may have also inadvertently led to a minimization of the magnitude of benefits of selective COX-2 inhibitors in reducing NSAID-related GI complications. Further research is needed to find better predictors for individual susceptibility to cardiovascular toxicity of selective COX-2 inhibitors.
We found that the incidence rate of upper GI events during therapy was approximately 0.4 cases per 100 person-years in patients receiving selective COX-2 inhibitors. In the 2 large clinical trials that demonstrated the efficacy of selective COX-2 inhibitors in reducing the risk of GI complications, the incidence rate of upper GI events was approximately 2 cases per 100 person-years in both the rofecoxib group and the celecoxib group (1, 2). Although there were several differences in the case definitions between these clinical trials and the GPRD and major differences in patient characteristics (e.g., age), it is possible that the patients included in the clinical trials for selective COX-2 inhibitors were more severely sick than those treated in general medical practice. The implication may be that the number needed to treat in order to prevent 1 NSAID-related GI complication may differ between randomized controlled trials and general medical practice. In addition, the benefit of prescribing selective over nonselective COX-2 inhibitors may be smaller in actual clinical practice, with a lower incidence of upper GI events. A recent study reported that the benefit of selective COX-2 inhibitors in reducing the frequency of upper GI events may be offset by their cardiovascular harm (12).
Our study found that approximately half of the patients prescribed selective COX-2 inhibitors did not have any major risk factors for upper GI events. This finding is similar to that reported for a Dutch population, which included similar GI risk factors (4). A US study that used survey data on patient visits also found similar results and reported that ∼60% of the patients prescribed selective COX-2 inhibitors could be classified as having a low risk for adverse GI events associated with NSAIDs (5). A French study reported that only 33% of patients receiving selective COX-2 inhibitors had a history of upper GI events or dyspepsia or GI reflux, and that almost half had a history of ischemic events or ischemic risk factors (6). Another UK study reported that <1% of patients taking celecoxib or rofecoxib had a history of upper GI events at the start of therapy (13). In our study, the estimate for patients with upper GI history was also low but considerably higher than that reported in another UK study (10). The reason for this discrepancy is unclear, but higher rates in the GPRD compared with the database from the other UK study have also been reported for other outcomes (14).
There are several limitations to this study. First, we were unable to separate each of the contributions of the various sources of publicity following the withdrawal of rofecoxib. Second, we compared different periods of time. In analyses that evaluate patterns over time, other unaccountable changes may have taken place. Thus, any interpretation of cause and effect needs to be made with caution. We also could not determine the reason for the significant differences between the UK regions in the use of selective COX-2 inhibitors following the withdrawal of rofecoxib. Prescribing advice given to practitioners concerning the use of selective COX-2 inhibitors may have differed between England, Wales, Northern Ireland, and Scotland.
In conclusion, there was no channeling in the usage of selective COX-2 inhibitors toward patients with high risk of upper GI events and low risk of cardiovascular disease following the withdrawal of rofecoxib.
Dr. van Staa had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Leufkens, van Staa.
Acquisition of data. Setakis, van Staa.
Analysis and interpretation of data. Setakis, Leufkens, van Staa.
Manuscript preparation. Leufkens, van Staa.
Statistical analysis. Setakis, van Staa.
We thank Dr. Lesley Wise for reviewing this article.