Retrospective analysis of utilization patterns and cost implications of coxibs among seniors in Quebec, Canada: What is the potential impact of the withdrawal of rofecoxib?

Authors

  • Elham Rahme,

    Corresponding author
    1. McGill University and Research Institute, McGill University Health Centre, Montreal, Quebec, Canada
    • Division of Clinical Epidemiology, Montreal General Hospital, 1650 Cedar Avenue, Montreal, Quebec H3G 1A4, Canada
    Search for more papers by this author
    • Dr. Rahme has received grants from, and has served as a consultant to, Merck & Co., Inc., the previous manufacturer of rofecoxib, and Pfizer, Inc., the manufacturer of celecoxib. On September 30, 2004, Merck & Co., Inc., announced the voluntary worldwide withdrawal of rofecoxib from the market.

  • Elke Hunsche,

    1. Outcomes Research, Merck & Co. Inc., Whitehouse Station, New Jersey
    Search for more papers by this author
  • Youssef Toubouti,

    1. McGill University and Research Institute, McGill University Health Centre, Montreal, Quebec, Canada
    Search for more papers by this author
  • Isabelle Chabot

    1. Health Economics and Outcomes Research, Merck Frosst Canada Ltd., Montreal, Quebec, Canada
    Search for more papers by this author

  • This article was prepared with the assistance of BioMedCom Consultants Inc., Montreal, Canada.

Abstract

Objective

In September 2004, the manufacturer of rofecoxib announced a voluntary worldwide withdrawal of the drug. The impact of this withdrawal on drug budgets is unclear. This study evaluated average daily doses and costs of rofecoxib and celecoxib and concomitant use of gastroprotective agents (GPAs) in elderly patients with osteoarthritis (OA) or rheumatoid arthritis (RA) in Quebec, prior to the rofecoxib withdrawal.

Methods

This retrospective cohort study used prescription drug and medical service data from the Quebec government health agency administrative database and included coxib users ≥66 years of age with OA or RA who filled ≥3 consecutive rofecoxib or celecoxib prescriptions in 2001–2002. Results were adjusted for gastrointestinal risk factors and other patient baseline characteristics.

Results

Data were analyzed for 11,975 rofecoxib and 12,480 celecoxib users. Mean daily dosages were 20.7 mg for rofecoxib and 231.3 mg for celecoxib. Rofecoxib users consumed a mean ± SD of 0.95 ± 0.43 pills per day, and celecoxib users took 1.34 ± 0.65 pills per day. Mean ± SD unadjusted daily acquisition costs were $1.18 ± $0.53 (Canadian) for rofecoxib and $1.45 ± $0.74 for celecoxib. After adjusting for patient baseline characteristics, the mean daily acquisition cost for rofecoxib was $0.25 lower than for celecoxib. Rofecoxib users were less likely than celecoxib users to fill a GPA coprescription (odds ratio 0.88; 95% confidence interval 0.81, 0.95). Subgroup analyses yielded comparable results.

Conclusion

Celecoxib appears to be a more expensive therapeutic option than rofecoxib due to a relatively higher daily dose and tablet consumption.

INTRODUCTION

Arthritis is a major cause of severe long-term pain and disability worldwide. Globally, an estimated 10% of men and 18% of women ages ≥60 years have osteoarthritis (OA), whereas rheumatoid arthritis (RA) affects 0.3–1% of the population (1). The arthritis prevalence in Canada is consistent with global trends; ∼10% of adults are afflicted with OA and 1% with RA (2). The economic burden of arthritis is high, with total costs in western nations representing 1–2.5% of the gross national product (3). Total costs of arthritis were conservatively estimated at $4.4 billion (unless otherwise specified, all dollar values are in Canadian dollars) in Canada in 1998, of which direct medical costs accounted for 20% and indirect costs due to lost productivity accounted for 80% (2). The average total costs of arthritis per patient were estimated to be ∼$11,000 for RA and $7,000 for OA in the province of Ontario in 2000 (4).

With no cure currently available for either OA or RA, the primary goal of therapy is to control pain. To this end, clinical practice guidelines for OA and RA recommend the use of nonsteroidal antiinflammatory drugs (NSAIDs) either as first-line therapy or following a trial of acetaminophen (5–8). However, nonselective NSAIDs have long been recognized to cause clinically significant upper gastrointestinal (GI) side effects, particularly among elderly adults and other high-risk patients (9–11). Based on estimates for the period between 1993 and 1997, the average cost per patient for treating GI events attributable to nonselective NSAIDs was $0.84–$0.94 per day of NSAID therapy in the province of Quebec (12, 13).

Risk factors for nonselective NSAID-related gastropathy include a history of peptic ulcer disease, concurrent use of anticoagulants or corticosteroids, cardiovascular comorbidity, and advanced age (14). With the aim of reducing the risk of upper GI adverse events, guidelines recommend that use of nonselective NSAIDs in patients at risk for GI events should be accompanied by prophylactic use of gastroprotective agents (GPAs) (5–8, 15, 16). Alternatively, these guidelines recommend the use of cyclooxygenase 2–specific inhibitors (coxibs), which have been demonstrated in large randomized, controlled trials to possess efficacy similar to maximum recommended doses of nonselective NSAIDs, but at a significantly lower risk of GI side effects (17–20).

In Quebec, the government health agency, the Régie de l'Assurance Maladie du Québec (RAMQ), has provided unrestricted reimbursement coverage of celecoxib since October 1999. RAMQ also provided unrestricted coverage of rofecoxib from April 2000 until September 2004 when the manufacturer of the drug announced a voluntary worldwide withdrawal. The acquisition cost of rofecoxib was $1.25 per 12.5-mg or 25-mg tablet, whereas celecoxib cost $0.625 per 100-mg capsule or $1.25 per 200-mg capsule. It can be expected that some of the patients who were using rofecoxib will be switched to celecoxib.

Differences in the reimbursed costs for the coxibs can be driven by differences in the dose strength per unit (tablet) prescribed for rofecoxib and celecoxib, the cost per unit, and the daily utilization by patients, which may well differ from the dose prescribed by the physician. Furthermore, because patients may discontinue coxib therapy and not refill their prescription (21, 22), it is important to consider prescriptions that are actually paid for when calculating expenditures for coxibs. Refilling of a prescription may be an indication that the patient has consumed the preceding dispensation from the same drug. Therefore, when calculating the drug budget impact, it is important to focus on the actual daily dose used by the patient and the actual cost incurred by the payer.

An additional potential source of variation in the overall drug cost of coxib therapy is the rate of GPA coprescription. Although GPAs are frequently prescribed for conditions such as gastroesophageal reflux disease (23), they are also used for prophylaxis and treatment of upper GI symptoms in nonselective NSAID users. A recent US study revealed that 42% of arthritis patients receiving nonselective NSAIDs also used GPAs: sucralfate, misoprostol, proton-pump inhibitors (PPIs), or histamine-2 receptor antagonists (H2RAs) (24). In Quebec, patients who took coxibs in 1999 and 2000 were 61% less likely to receive a GPA coprescription than those who took nonselective NSAIDs (25). However, it has not previously been reported whether the rate of GPA coprescription, and thus GPA expenditure, differs between coxibs.

The objective of this study was to compare the actual use of rofecoxib and celecoxib by patients with OA or RA as measured by average daily dose, average daily drug acquisition cost, and concomitant use of GPAs in the Quebec elderly adult population. The study design provides a basis for assessing whether any changes in coxib reimbursement costs can be expected after the withdrawal of rofecoxib.

PATIENTS AND METHODS

Study design.

This was a retrospective cohort study using the RAMQ pharmaceutical and medical services administrative database for the period January 2001 to December 2002. As in other Canadian provinces, all Quebec residents are eligible for public insurance coverage of physician visits and medical services provided by clinics and hospitals. RAMQ also covers the cost of prescription drugs for residents ≥65 years of age, welfare recipients, and those without access to private drug plan coverage, such as individuals who are self employed. Data in the RAMQ database have been validated previously and found to be accurate and reliable (26).

Demographic, prescription drug, and medical services data were extracted for senior adults who filled at least 3 consecutive prescriptions for either rofecoxib or celecoxib, with the first of these filled between January 1 and December 31, 2001. The index date for each patient in the study was the dispensing date of the first of the 3 consecutive prescriptions. Each patient was assigned to 1 of 2 study groups, rofecoxib or celecoxib, based on the coxib dispensed at the index date. The followup period lasted from the index date to the dispensing date of the third prescription, thereby capturing utilization data for the first 2 consecutive prescriptions. Requiring patients to fill a third prescription was expected to increase the likelihood that the pills supplied for the first 2 prescriptions were actually consumed during the followup period.

Patients.

Only RAMQ enrollees at least 66 years old were included in the study. The minimum age of 66 years was necessary to ensure that at least 1 year of prior drug utilization data would be available for each patient. Patient inclusion criteria were as follows: chronic coxib use defined as at least 3 consecutive prescriptions filled for either rofecoxib or celecoxib beginning in 2001, representing ≥60 days of treatment in the year following the index date. Consecutive coxib prescriptions were those with no intervening prescriptions for a nonselective NSAID (including meloxicam) or acetaminophen, and a refill interval ≤90 days between the date on which the supply of one prescription ended (prescription duration was included in the database) and the refill date of the next. Patients who switched between coxibs during the followup period were excluded. The analysis was restricted to patients with a diagnosis of OA or RA in the year prior to the index date on the basis of International Classification of Diseases, Ninth Edition codes (ICD-9; codes 715.x or 714.x, respectively) (27).

Patient characteristics assessed at the index date included risk factors for NSAID-related gastropathy as well as indicators of past medical attention for GI conditions. Drug use history was assessed by prescriptions filled in the year prior to the index date for a GPA (PPI, H2RA, or misoprostol), coxib, nonselective NSAID, acetaminophen, opiate, or antihypertensive agent. In addition, concomitant prescriptions for an anticoagulant, acetylsalicylic acid (ASA), corticosteroid, or GPA filled at the index date or before the index date with sufficient days of supply to overlap the index date were evaluated. Prescriptions for GPAs filled at the index date (not before) were evaluated separately; these were considered more likely to be for prophylactic use than prescriptions filled before the index date, which could have been for treatment of prior GI conditions. Medical history variables in the prior year included a diagnosis of gastric perforation, ulcer, or bleeding (PUB; ICD-9 codes 531.x–534.x and 578.x); nonbenign cancer (ICD-9 codes 140.x–208.x and 230.x–239.x); renal failure (ICD-9 codes 584.x–586.x); upper GI investigation (endoscopy or barium sulfate test); and hospitalizations with GI-related diagnoses (primary or secondary discharge diagnosis of stomach or duodenal PUB). Additional variables examined were age, sex, visits to rheumatologists in the prior year, and ischemic heart disease (ICD-9 codes 410.x–414.x) in the prior year.

Study endpoints.

The outcomes evaluated in this study were daily consumption and cost of coxibs, and coprescription of GPAs with coxibs. Consumption of coxibs was calculated both as the average daily dose (mg) and the average daily number of pills, reflecting patient use rather than prescribed doses. It was assumed that all pills supplied in the first 2 filled prescriptions were consumed between the index date and the dispensing date of the third prescription. The average daily consumption was calculated as follows:

equation image

and

equation image

The average daily acquisition cost for coxibs was calculated from a single payer perspective: rofecoxib, $1.25 per 12.5-mg or 25-mg tablet; celecoxib, $0.625 per 100-mg capsule or $1.25 per 200-mg capsule. Total acquisition cost did not include pharmacist dispensing fees. Average daily acquisition cost was calculated as follows:

equation image

Coprescription of GPAs at the index date was also compared between the rofecoxib and celecoxib groups.

Statistical analysis.

Patient characteristics, average daily doses, acquisition costs, and number of pills were analyzed by descriptive statistics. Logistic regression was used to compare patient characteristics at the index date between the rofecoxib and celecoxib study groups. The average daily acquisition costs of the 2 groups were compared by linear regression analysis, adjusting for patient characteristics at the index date. GPA use at the index date was compared between the rofecoxib and the celecoxib groups by logistic regression. This was a dichotomous variable, scored as whether or not a PPI, H2RA, or misoprostol was dispensed at the index date.

All regression models used a stepwise procedure to evaluate the explanatory significance of the following variables: age at the index date; sex; concomitant prescriptions for corticosteroids, anticoagulants, and ASA; prescriptions filled in the year preceding the index date for GPAs, antihypertensive agents, acetaminophen, NSAIDs, and opiates; visits in the preceding year to rheumatologists and gastroenterologists; GI hospitalization and upper GI investigations in the preceding year; and diagnoses in the preceding year for OA versus RA, PUB, ischemic heart disease, cancer, and renal failure. Potential explanatory variables were assessed for colinearity by examining the variance inflation factor. A variance inflation factor >10 was considered to indicate colinearity (28), and was a criterion for variable exclusion. Only independent variables with P < 0.10 were retained in the regression models.

Subgroup analyses were performed as sensitivity analyses. In the first subgroup analysis, potential confounding of results by including both new and continuous coxib users was addressed by assessing outcomes for coxibnaïve patients only, excluding those who filled coxib prescriptions in the year preceding the index date. The second subgroup analysis reduced the heterogeneity of the cohort by assessing outcomes for patients with OA only, excluding those diagnosed with RA. Regression models for the subgroup analyses included all potential explanatory variables that were retained in the models for the main analyses.

RESULTS

Patient characteristics at the index date.

Data were analyzed for 24,455 RAMQ enrollees who met the inclusion and exclusion criteria. Most patient characteristics were comparable between coxib groups (Table 1). No evidence was found of colinearity between these variables. As shown in Table 2, the proportion of patients with prior GI hospitalizations was numerically but not significantly higher in the rofecoxib group. The odds of having concomitant corticosteroid prescriptions at or overlapping the index date were 23% lower in the rofecoxib group than in the celecoxib group. The odds of having been diagnosed with RA were 35% lower in the rofecoxib group than in the celecoxib group. Although there was a statistically significant trend for patients in the rofecoxib group to be younger than those in the celecoxib group (odds ratio [OR] 0.99; 95% confidence interval [95% CI] 0.99, 0.99), the difference in average age was only 0.4 years and is unlikely to be of importance for this analysis.

Table 1. Patient characteristics at the index date*
CharacteristicRofecoxibCelecoxib
  • *

    The index date is the date on which the first coxib prescription in the followup period was filled. Values are the percentage unless otherwise indicated. ASA = acetylsalicylic acid; NSAID = nonsteroidal antiinflammatory drug; GI = gastrointestinal; GPA = gastroprotective agent; PUB = perforation, ulcer, or bleeding.

  • Concomitant prescriptions were filled on the index date or before the index date with days of supply of medication overlapping the index date.

Number of patients11,97512,480
Age, mean ± SD years75.1 ± 5.975.5 ± 5.9
Women72.572.5
Concomitant prescriptions  
 Corticosteroids4.36.6
 Anticoagulants2.62.8
 ASA23.523.5
Prescriptions filled in preceding year  
 Antihypertensive agents48.949.6
 Acetaminophen43.342.4
 Nonselective NSAIDs24.717.0
 Opiates7.16.0
 Rofecoxib73.916.5
 Celecoxib40.182.1
Visit(s) to rheumatologists in preceding year21.023.0
Diagnoses in preceding year  
 Rheumatoid arthritis11.717.2
 Osteoarthritis92.687.4
 Ischemic heart disease17.216.5
 Cancer12.211.9
 Renal failure1.31.2
GI-related events in preceding year44.844.2
 GPA prescription (filled)41.340.8
 Visits to gastroenterologists8.88.1
 Upper GI investigations6.05.5
 Diagnosis of PUB1.61.4
 GI hospitalization0.60.5
Concomitant GPA prescription28.129.0
GPA prescription filled on the index date14.615.8
Table 2. Comparison of patient baseline characteristics associated with the prescription of rofecoxib versus celecoxib*
CharacteristicAll patientsNew coxib usersOA patients
  • *

    Values are the odds ratio (95% confidence interval) based on logistic regressions that assessed all variables presented in Table 1 except the following: rofecoxib or celecoxib prescriptions in the preceding year, concomitant GPA prescriptions, and GPA coprescriptions filled at the index date. GI-related events in the preceding year were considered separately and not as a group. The regressions retained only the explanatory variables listed in this table. OA = osteoarthritis; RA = rheumatoid arthritis; NSAID = nonsteroidal antiinflammatory drug; GI = gastrointestinal.

  • Patients who filled prescriptions for a coxib in the year prior to the index date are excluded.

  • Patients with a diagnosis of rheumatoid arthritis are excluded.

Age, 1-year increase0.99 (0.99, 0.99)1.00 (0.99, 1.01)0.99 (0.98, 0.99)
RA diagnosis in year preceding index date0.65 (0.60, 0.70)0.75 (0.61, 0.92) 
Concomitant corticosteroids0.77 (0.68, 0.87)0.84 (0.61, 1.16)0.83 (0.69, 1.00)
Prescriptions in year preceding index date   
 Nonselective NSAIDs1.62 (1.52, 1.72)1.03 (0.89, 1.20)1.54 (1.44, 1.65)
 Opiates1.18 (1.06, 1.31)1.27 (0.95, 1.70)1.18 (1.05, 1.32)
GI hospitalization1.33 (0.95, 1.87)1.33 (0.58, 3.05)1.24 (0.85, 1.80)

Coxib use and costs.

Coxib exposure was comparable in the rofecoxib and celecoxib groups, whether measured as the number of days of pills supplied at the first 2 prescriptions (mean ± SD [median] 58.5 ± 13.4 [60] and 59.1 ± 12.3 [60], respectively) or the number of days between the index and third dispensing dates (mean ± SD [median] 76.8 ± 32.7 [64] and 78.8 ± 33.5 [64], respectively). As shown in Table 3, the mean daily rofecoxib dosage, ∼21 mg, was below the maximum of 25 mg that is recommended for chronic treatment of both OA and RA. The mean daily celecoxib dosage, ∼231 mg, was between the maximum recommended dosages for OA (200 mg) and RA (400 mg). On average, patients in the rofecoxib group took 0.39 fewer pills daily than did patients in the celecoxib group. The average unadjusted daily cost was $0.27 (23%) higher in the celecoxib group than in the rofecoxib group.

Table 3. Unadjusted mean ± SD coxib use and daily costs
 All patientsNew coxib users*Osteoarthritis patients
RofecoxibCelecoxibRofecoxibCelecoxibRofecoxibCelecoxib
  • *

    Patients who filled prescriptions for a coxib in the year prior to the index date are excluded.

  • Patients with a diagnosis of rheumatoid arthritis are excluded.

  • Maximum recommended daily doses: rofecoxib, 25 mg for osteoarthritis or rheumatoid arthritis; celecoxib, 200 mg for osteoarthritis, 400 mg for rheumatoid arthritis.

Average daily dose (mg)20.7 ± 10.2231.3 ± 118.219.65 ± 10.71206.18 ± 107.5920.45 ± 10.13220.28 ± 109.59
Average number of pills per day0.95 ± 0.431.34 ± 0.650.91 ± 0.451.19 ± 0.640.93 ± 0.421.29 ± 0.62
Average daily acquisition cost ($)1.18 ± 0.531.45 ± 0.741.14 ± 0.561.29 ± 0.671.17 ± 0.531.38 ± 0.69

Daily acquisition costs for coxibs were also compared after adjusting for explanatory variables using linear regression. The adjusted average daily acquisition cost of rofecoxib was significantly lower than that of celecoxib by $0.25 (Table 4). Adjusted daily coxib costs were $0.17 higher for patients with RA than for those with OA, and patients who visited rheumatologists in the preceding year incurred $0.12 higher adjusted daily coxib expenditure than those who did not visit rheumatologists. On average, women had slightly ($0.03) but significantly lower adjusted daily coxib costs than men.

Table 4. Comparison of daily acquisition costs for coxibs after adjusting for patient baseline characteristics*
Patient characteristicAdjusted difference in daily acquisition cost in $ (95% CI)
All patientsNew patientsOA patients§
  • *

    95% CI = 95% confidence interval; OA = osteoarthritis; vs = versus; see Table 1 for additional definitions.

  • Based on linear regressions that assessed all variables presented in Table 1 except the following: rofecoxib or celecoxib prescriptions in the preceding year, concomitant GPA prescriptions, and GPA coprescriptions filled at the index date. GI-related events in preceding year were considered separately and not as a group. The regressions retained only the explanatory variables listed in this table.

  • Patients who filled prescriptions for a coxib in the year prior to the index date are excluded.

  • §

    Patients with a diagnosis of rheumatoid arthritis are excluded.

Rofecoxib vs celecoxib−0.25 (−0.27, −0.24)−0.14 (−0.17, −0.11)−0.21 (−0.23, −0.20)
Rheumatoid arthritis vs osteoarthritis0.17 (0.15, 0.20)0.18 (0.12, 0.24) 
Women vs men−0.03 (−0.05, −0.02)−0.02 (−0.05, 0.01)−0.04 (−0.05, −0.02)
Visits to rheumatologists in preceding year (any vs none)0.12 (0.10, 0.14)0.08 (0.02, 0.13)0.10 (0.08, 0.12)
Concomitant prescriptions (any vs none)   
 Corticosteroids0.14 (0.11, 0.17)0.12 (0.03, 0.20)0.15 (0.10, 0.20)
 ASA0.03 (0.02, 0.05)0.03 (0.00, 0.07)0.04 (0.02, 0.06)
Prescriptions filled in preceding year (any vs none)   
 GPA0.07 (0.05, 0.08)0.04 (0.00, 0.07)0.07 (0.05, 0.09)
 Nonselective NSAID0.07 (0.05, 0.09)0.10 (0.06, 0.14)0.07 (0.05, 0.09)
 Acetaminophen0.09 (0.07, 0.10)0.06 (0.02, 0.09)0.10 (0.08, 0.12)
 Antihypertensive agent0.05 (0.03, 0.06)0.02 (−0.02, 0.05)0.06 (0.04, 0.07)
GI hospitalizations in preceding year (any vs none)0.10 (0.01, 0.20)0.25 (0.04, 0.47)0.12 (0.01, 0.22)

Patients who received concomitant corticosteroid or ASA therapy had significantly higher adjusted daily coxib costs than those who did not. Similarly, patients who had filled prescriptions for GPAs, nonselective NSAIDs, acetaminophen, or antihypertensive agents in the year preceding the index date had significantly higher adjusted daily coxib costs than those who had not. GI hospitalization in the prior year was also associated with significantly higher adjusted daily coxib costs.

Gastroprotective agent use at the index date.

After adjusting for explanatory variables using logistic regression, the odds of filling a coprescription for a PPI, H2RA, or misoprostol at the index date were 12% lower in the rofecoxib group than in the celecoxib group (Table 5). The odds of filling a GPA coprescription were 15% higher for patients with RA than those with OA. In addition, patients with prescriptions for anticoagulants or ASA at or overlapping the index date had a significantly higher likelihood of filling a GPA coprescription than did those without concomitant use of these medications. Patients had >28 times greater odds of receiving GPAs at the index date if they had taken them in the preceding year. The odds of filling a GPA coprescription at the index date were 57% higher for patients hospitalized in the year prior to the index date with a GI-related primary or secondary discharge diagnosis than for those with no GI hospitalization. There were no significant associations between GPA coprescription and other baseline characteristics.

Table 5. Comparison of gastroprotective agent (GPA) use at the index date after adjusting for patient baseline characteristics*
Patient characteristicOR (95% CI)
  • *

    OR = odds ratio; 95% CI = 95% confidence interval; vs = versus; see Table 1 for additional definitions.

  • Based on logistic regression that assessed all variables presented in Table 1 except the following: rofecoxib or celecoxib prescriptions in the preceding year, concomitant GPA prescriptions, and GPA coprescriptions filled at the index date. GI-related events in the preceding year were considered separately and not as a group. The regression retained only the explanatory variables listed in this table.

Rofecoxib vs celecoxib0.88 (0.81, 0.95)
Rheumatoid arthritis vs osteoarthritis1.15 (1.04, 1.28)
Women vs men1.08 (0.99, 1.19)
Concomitant anticoagulants (any vs none)1.30 (1.04, 1.63)
Concomitant ASA (any vs none)1.27 (1.16, 1.39)
GPAs in the preceding year (any vs none)28.31 (24.85, 32.25)
GI hospitalizations in preceding year (any vs none)1.57 (1.09, 2.24)

Subgroup analyses.

When analyses were restricted to coxib-naïve patients (excluding patients with coxib prescriptions in the year prior to the index date), logistic regression revealed that patients in the rofecoxib group were 25% less likely to have RA than those in the celecoxib group (Table 2). Other patient baseline characteristics did not differ significantly between the 2 coxib groups. The unadjusted average daily rofecoxib dosage was ∼20 mg (Table 3), similar to that in the primary analysis, but the average daily celecoxib dosage, 206 mg, was lower. Table 4 shows that after adjusting for explanatory variables, the daily acquisition cost of rofecoxib remained $0.14 lower than that of celecoxib in coxib-naïve patients. In this scenario, there was no difference between rofecoxib and celecoxib regarding the likelihood of GPA dispensing at the index date (OR 0.99; 95% CI 0.80, 1.22).

If only patients with OA were considered (excluding patients with an RA diagnosis in the prior year), the differences between rofecoxib and celecoxib groups in age, concomitant corticosteroid use, and prior prescriptions for nonselective NSAIDs and opiates were similar to those in the primary analysis (Table 2). Other patient baseline characteristics did not differ significantly between the coxib groups. The unadjusted average daily rofecoxib dosage was ∼20 mg whereas the average daily celecoxib dosage was 220 mg (Table 3). The adjusted daily acquisition cost of rofecoxib was $0.21 lower than that of celecoxib (Table 4). Fewer patients with OA in the rofecoxib group than in the celecoxib group received a GPA at the index date (OR 0.86; 95% CI 0.79, 0.94).

DISCUSSION

This study of patients with OA and RA ages ≥66 years demonstrated that the average adjusted daily costs were significantly lower for rofecoxib than for celecoxib by $0.25 per patient. This difference resulted from a relatively lower daily dose and tablet consumption of rofecoxib than celecoxib. On average, patients with arthritis in this cohort who received rofecoxib used less than the 25-mg maximum daily dosage recommended for OA or RA, whereas those who received celecoxib used more than the 200-mg maximum dosage recommended for OA but less than the 400-mg maximum dosage recommended for RA.

This difference in daily costs cannot be explained by the higher percentage of celecoxib patients with RA. For this condition, the maximum recommended daily celecoxib dosage (400 mg) is twice as high as the recommended dosage for OA, but for rofecoxib it is the same for RA and OA (25 mg). Nevertheless, even when patients with a diagnosis of RA were excluded from the analysis, the adjusted daily acquisition cost of rofecoxib remained $0.21 lower than that of celecoxib.

Recent studies in other regions that were completed before the withdrawal of rofecoxib yielded similar findings. Two of these studies were based on US pharmacy claims databases, although differences in methodology confound direct comparisons. Schnitzer et al (29) conducted a database analysis of >220,000 prescriptions for >58,000 chronic coxib users in the US (medication supply ≥90 days between June 1, 2000 and May 31, 2001). They calculated the mean prescribed daily dose, i.e., the total supplied dose divided by the number of days of supply if medications were taken as prescribed (29). The mean rofecoxib dosages for patients with OA and RA were 28.2 mg/day and 29.0 mg/day, respectively. The mean celecoxib dosages were 267.9 mg/day and 324.0 mg/day, respectively. Treatment with rofecoxib was less expensive than with celecoxib due to the lower number of rofecoxib tablets per day. Based on average wholesale prices (AWP), a day's treatment with rofecoxib would have cost an estimated $3.40 (US dollars) for OA and $3.49 for RA, compared with daily costs of $3.86 and $4.67, respectively, for celecoxib (29). Because this study was based on prescribed doses rather than doses used by the patient, and because patient populations were different, comparison of results is limited.

Mean daily doses and mean daily costs of coxib prescriptions dispensed are also available from a study by Weideman et al that was based on the US Veterans Affairs (VA) system in 2000 (30). Using total days supplied if coxibs were taken as prescribed as the measure of treatment duration, this study found average daily dosages of 237 mg for celecoxib and 24 mg for rofecoxib. Drug prices in the VA system are often much lower than the AWP. Average daily costs were $1.57 (US dollars) for celecoxib and $1.37 for rofecoxib.

Furthermore, a recently presented study of coxib therapy for arthritis in Norway determined an average adjusted daily cost for OA therapy that was 0.39 Norwegian Krone (NOK; 1 NOK = $0.20 Canadian dollars in 2002) lower for rofecoxib than for celecoxib, for the period November 2001 to December 2002 (31). Average daily costs for RA therapy were 2.05 NOK lower for rofecoxib than for celecoxib.

As suggested by Schnitzer et al (29) and Weideman et al (30), differences in the pharmacokinetics of rofecoxib and celecoxib may underlie differences in the average daily doses of these drugs that are consumed. Because the half life of rofecoxib is ∼17 hours whereas that of celecoxib is ∼11 hours, the recommended dosing for rofecoxib is 25 mg once daily, whereas the recommended dosing for celecoxib is 100–200 mg twice daily for RA and 200 mg once daily or 100 mg twice daily for OA (32). This difference in recommended dosing might explain the lower average daily number of pills used by rofecoxib recipients in this study cohort compared with celecoxib recipients (0.95 versus 1.34 pills per day).

The total daily drug costs of coxibs are also impacted by the prophylactic use of GPAs (33). In this study of coxib recipients, patients who received rofecoxib were 12% less likely to fill a GPA coprescription at the index date than were those who received celecoxib. This may have translated into lower average daily costs of arthritis treatment with rofecoxib than with celecoxib to the provincial drug program. However, no conclusion regarding the association between the prescription of GPAs and the use of coxibs can be drawn from this study because the medical reasons for GPA use were unknown.

The results of this study are subject to some limitations, many of which are inherent to this type of analysis. Although regression models were used to adjust for several potentially confounding patient characteristics, results may be biased by other variables that were unaccounted for (i.e., those not recorded in the RAMQ database, such as disease severity or obesity). Selection bias cannot be excluded because physicians' reasons for prescribing the medications of interest in this study were not available (34). Daily doses used by patients were calculated based on refill intervals in the absence of compliance data and detailed daily drug consumption information. However, the average daily costs that were calculated reflect actual costs incurred, and are therefore of relevance to payers.

Another limitation is that the RAMQ database provides no information on use of over-the-counter medications in the study cohort. However, GPAs, coxibs, and nonselective NSAIDs other than ASA and ibuprofen are only available by prescription in Canada, and patients have a financial incentive to obtain drugs by prescription because the prescription is covered by RAMQ.

The prescriptions evaluated in this study did not represent all coxib use by elderly patients in Quebec. Although a gap of 90 days was permitted between prescription refills, most patients filled their prescriptions much more quickly. The median duration of coxib supply for the first 2 prescriptions was 60 days and the median time between the first and third prescriptions was 64 days. This suggests that the patients experienced relatively severe pain; therefore, these results may not apply to patients with mild symptoms who may use coxibs more sporadically or at lower doses. In addition, the study design did not permit switching between coxibs, and may reflect costs only among patients who were satisfied with their therapy. However, in another study based on the RAMQ database for the same period, we found a low rate of switching between coxibs or switching to a nonselective NSAID among patients ≥65 years of age: 3.2% switched from rofecoxib to either celecoxib or a nonselective NSAID, and 3.4% switched from celecoxib to either rofecoxib or a nonselective NSAID (35).

Including other therapy-related expenditures could influence the relative costliness of these drugs if adverse event rates differed between rofecoxib and celecoxib. However, it is not currently possible to conduct a rigorous comparison of the costs of adverse events associated with coxib therapy because no long-term randomized clinical outcome trials providing head-to-head comparisons of GI or cardiovascular risks of coxibs have been published to date. Furthermore, the available observational studies have intrinsic limitations and provide inconclusive results. For example, one retrospective database study found a significantly higher rate of hospitalization for upper GI hemorrhage among rofecoxib users than celecoxib users (36), but trends towards a higher rate of GI adverse events for rofecoxib than celecoxib in 2 retrospective case-control studies appeared not to be significant (37, 38). Some observational studies found significantly higher rates of various serious cardiovascular events for rofecoxib compared with celecoxib (39–44), whereas other studies indicated no significant difference (45–47). The US Food and Drug Administration recently concluded that currently available data do not allow ranking of coxibs with regard to cardiovascular risk (48).

In conclusion, for elderly patients with arthritis receiving chronic coxib treatment, the estimated cost of antiinflammatory therapy was $0.25 lower per patient per day for rofecoxib than for celecoxib due to a relatively lower daily dose and tablet consumption. Additional costs can be expected due to the greater concomitant use of GPAs among celecoxib users than rofecoxib users. Celecoxib therefore appears to be a more expensive therapeutic option than rofecoxib in this patient group. If at least some patients who have previously received rofecoxib switch to celecoxib, the withdrawal of rofecoxib might increase overall expenditure on chronic antiinflammatory therapy for patients with OA or RA. However, the magnitude of this impact will depend on how many former rofecoxib users switch to celecoxib or to nonselective NSAIDs or other analgesics, the concomitant utilization of GPAs, and other treatment-related cost differences between therapies.

Ancillary