Skip Navigation


Rheumatology Advance Access originally published online on February 3, 2006
Rheumatology 2006 45(7):903-910; doi:10.1093/rheumatology/kei269
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
45/7/903    most recent
kei269v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (3)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Rahme, E.
Right arrow Articles by Hunsche, E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rahme, E.
Right arrow Articles by Hunsche, E.
Related Collections
Right arrow Health Economics
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2006. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

Therapy switching and associated costs in elderly patients receiving COX-2 selective inhibitors or non-selective non-steroidal anti-inflammatory drugs in Quebec, Canada

E. Rahme, Y. Toubouti and E. Hunsche1

Department of Medicine, McGill University and Division of Clinical Epidemiology, Research Institute, McGill University Health Centre, Montreal General Hospital Site, Montreal, QC, Canada and 1 Outcomes Research Merck & Co., Inc., Whitehouse Station, NJ, USA.

Correspondence to: E. Rahme, Division of Clinical Epidemiology, Montreal General Hospital, Suite L10-408, 1650 Cedar Ave Montreal, Quebec H3G 1A4, Canada. E-mail: elham.rahme{at}mcgill.ca


   Abstract
Top
 Abstract
Introduction
 Methods
 Statistical analyses
 Results
 References
 
Objectives. Lack of efficacy or tolerability of some non-steroidal anti-inflammatory drugs (NSAIDs) may lead to switching between non-selective NSAIDs (nsNSAIDs) and cyclooxygenase-2 (COX-2) selective inhibitors (coxibs), potentially increasing treatment costs due to additional physician visits and wastage of medication. This study assessed drug switching and associated costs among elderly chronic NSAID users.

Methods. Data for patients who filled their first prescription for a coxib or nsNSAID in 2001 were obtained from the Quebec Health Insurance Agency. Follow-up was terminated at the earliest of: 1 yr, the first day without NSAID exposure following the index filling date, or death. Patients could switch NSAIDs several times during follow-up. Person-days of exposure were categorized by the NSAID most recently dispensed: rofecoxib, celecoxib, Arthrotec® or non-Arthrotec (nA) nsNSAID. Cox regression models compared time to switch between groups, adjusting for patient baseline characteristics. Upon a switch, pills remaining from the previous prescription were considered wasted. The costs of wasted pills and switch-associated physician visits were estimated.

Results. Throughout follow-up, patients filled 38 267 prescriptions for rofecoxib, 31 282 for celecoxib, 1108 for Arthrotec and 4388 for nA-nsNSAIDs. Adjusted hazard ratios (95% confidence interval) for switching versus nA-nsNSAIDs were: rofecoxib, 0.39 (0.35–0.44); celecoxib, 0.43 (0.38–0.48). Compared with nA-nsNSAID prescriptions, adjusted switching-related healthcare costs were 53 and 47% lower on average for rofecoxib and celecoxib prescriptions, respectively. These costs were 34% higher for Arthrotec prescriptions than for nA-nsNSAIDs.

Conclusions. Compared with recipients of nsNSAIDs, coxib recipients were less likely to switch medications and had approximately half the adjusted costs for switching-related wasted resources per prescription.

KEY WORDS: Cost analysis, Cyclooxygenase-2 inhibitors, Non-steroidal anti-inflammatory agents, Prescriptions


   Introduction
Top
 Abstract
 Introduction
Methods
 Statistical analyses
 Results
 References
 
Non-steroidal anti-inflammatory drugs (NSAIDs) are among the most frequently prescribed medications worldwide [1]. These agents are effective in reducing the pain and inflammation associated with chronic musculoskeletal conditions, but it has long been recognized that non-selective NSAIDs (nsNSAIDs) also increase the risk for gastrointestinal (GI) events [2–4]. Clinical trials have demonstrated that cyclooxygenase-2 (COX-2) selective inhibitors (coxibs) provide similar arthritis pain relief to nsNSAIDs prescribed at recommended doses while reducing the risk for upper and lower GI events compared with nsNSAIDs [5–8], although unlike rofecoxib, the GI benefit observed for celecoxib in short-term treatment [9] appears to diminish after 6 months of follow up [10].

Recent evidence suggests that both coxibs and nsNSAIDs (henceforth, the term nsNSAID refers only to non-aspirin nsNSAIDs) may increase the risk for cardiovascular (CV) adverse events [11–20]. In this environment, an increased rate of switching between different anti-inflammatory medications may be predicted as physicians alter their prescribing practice in response to new information. However, medication switching due to dissatisfaction with different NSAIDs was observed before concerns over their CV safety were widely disseminated. Published studies assessing nsNSAID utilization suggest a high level of switching, commonly related to lack of efficacy or tolerability [21–23]. Comparative studies have found lower switching rates with coxibs than nsNSAIDs [24–27].

Switching may result in wastage of medication and generate costs for additional physician visits. Therefore it is important to consider these costs when projecting the overall economic impact of a changing NSAID prescribing practice. The objective of this study was to identify determinants of switching among elderly new chronic users of coxibs or nsNSAIDs in 2001–2002 in Quebec, Canada, and to estimate the costs of healthcare resources associated with switching. The results of this study should be relevant to current circumstances, as the cost of a switch is expected to remain fairly stable over time.


   Methods
Top
 Abstract
 Introduction
 Methods
Statistical analyses
 Results
 References
 
Study design
This was a retrospective cohort study based on the Quebec government medical and pharmaceutical claims databases. All Quebec residents aged 65 yr or older (953 656 individuals in 2001) are eligible for coverage under the Provincial Health Care Fund administered by the Régie de l’Assurance Maladie du Québec (RAMQ) [28]. RAMQ has provided unrestricted coverage to celecoxib since October 1999 and to rofecoxib from April 2000 until its withdrawal in September 2004. This study was approved by the Ethics Review Board of the Montreal General Hospital.

Study population
The study cohort included beneficiaries 66 yr of age or older who filled an NSAID prescription in 2001. For each patient, the date of the first dispensation for any NSAID in 2001 was defined as the index date. The end of the study period was 31 December 2002.

Patients were included only if they were chronic users of coxibs or nsNSAIDs. Chronic use was defined as at least two consecutive NSAID dispensations in the year following the index date, with a total duration of medication supply of 60 days or more. There was no requirement that consecutive dispensations be for the same NSAID or within the same class of NSAID (coxib or nsNSAID). Two dispensations were considered consecutive if the second occurred during the period covering 1.25 times the intended therapy duration of the first (as recorded by RAMQ). As patients may not take NSAIDs as frequently as prescribed and in the absence of direct measurement of adherence in the database, we assumed NSAIDs were taken on 80% of prescribed days (reciprocal = 1.25), a commonly cited threshold for acceptable adherence to chronic therapy [29].

Only new users of NSAIDs were included to avoid the possibility that the index prescription represented a medication switch. Therefore patients who filled any NSAID prescription in the year preceding their index date were excluded.

To capture medical histories consistently, all subjects were required to have at least 365 consecutive days of information in the RAMQ database preceding the index date. Patients were followed until the earliest of 1 yr after the index date, the first non-exposure date (defined below) or death.

Baseline characteristics
Patient characteristics determined at the index date (listed in Table 1) included demographic variables, chronic disease diagnoses during the prior year and medications dispensed at the index date and during the prior year. Disease diagnoses were inferred from International Classification of Diseases, Ninth Edition (ICD-9) codes [30] and, in the case of diabetes and hypertension, also from use of disease-specific medications. Gastroprotective agents (GPAs)—misoprostol, proton pump inhibitors (PPIs) and histamine-2 receptor antagonists (H2RAs)—were assessed at the index date and during the prior year. Medications known to increase the risk for GI events (e.g. anticoagulants or corticosteroids) were assessed for the year preceding the index date. Concomitant aspirin use was determined as aspirin prescriptions dispensed at or prior to the index date with days of supply extending beyond the index date. Information on non-prescription aspirin use was not available, although the prevalence of use of over-the-counter (OTC) aspirin is known to be low in this population (approximately 2%) [31].


View this table:
[in this window]
[in a new window]
  		TABLE 1. Patient characteristics at the index date

 
Drug exposure and switching
The number of days supplied for an NSAID prescription was ascertained from the RAMQ database (prescribed intended therapy duration). If a prescription was refilled before the end of the intended therapy duration (1.25 x days supplied), the exposure was extended as of the switching date as 1.25 x days supplied for the new dispensation. If no NSAID prescription was filled before an intended exposure episode ended, follow-up was terminated on the first date without NSAID exposure.

Patients with a coxib exposure and subsequent nsNSAID exposure starting before 1.25 x days of supplied coxib were assumed to have stopped taking the coxib and switched to the nsNSAID. Therefore, the coxib exposure episode was terminated at the onset of nsNSAID exposure. The corresponding assumption was applied to patients with nsNSAID exposure and subsequent coxib exposure. Switching between different nsNSAIDs was not considered, with the exception of the combination product Arthrotec® (diclofenac plus misoprostol). Previous research has shown that the discontinuation rate for nsNSAIDs increases by 36% if misoprostol is added, due to GI symptoms associated with this GPA (primarily diarrhoea and dyspepsia) [32]. Consequently, we analysed Arthrotec users separately from users of all other nsNSAIDs.

Costs
Costs of switching-related healthcare resources associated with NSAID switching were calculated as the acquisition costs of wasted medication plus the costs of switch-associated physician visits. The previous definition of exposure episode was applied when calculating mean daily acquisition costs:

Formula
Wasted medication costs were calculated by multiplying the mean daily acquisition cost for the pre-switch NSAID by the number of days between the switch date and the date on which the exposure episode for the pre-switch prescription would have ended.

When defining exposure episodes, it was assumed that if a refill of the same drug was dispensed before the end of the current supply, the total number of pills was consumed between the filling date of the first prescription and the last day of the last prescription +25% of the days supplied for the last prescription.

The cost of a physician visit was the average reimbursed by RAMQ, differentiated by type of physician [i.e. general practitioner (GP) or specialist]. A switch was associated with a physician visit if the prescription filled at the switch date had been written during the visit. If the patient did not have a written script (e.g. medication was dispensed based on a phone conversation between the physician and pharmacist), then no visit cost was added because there is no charge for over-the-phone prescriptions. All costs are reported in unadjusted Canadian dollars (US$1.00 = Can$1.55 in 2001, Can$1.57 in 2002) [33].


   Statistical analyses
Top
 Abstract
 Introduction
 Methods
 Statistical analyses
Results
 References
 
First switch
Patients were categorized into four groups—rofecoxib, celecoxib, Arthrotec and other nsNSAIDs—according to the drug dispensed at the index date. Descriptive statistics were used to compare patient baseline (at index date) characteristics. Crude rates of first switch were calculated for each group by dividing the number of patients who switched medication by the total number of person-days from index date to the first switch or the end of follow-up for patients who did not switch medications. A Cox regression model was used to compare hazard ratios (HRs) for first switch adjusting for baseline patient characteristics. Exposure groups were forced to be included in the model while other independent variables were kept in the model if statistically significant at {alpha} = 0.10 or if exclusion caused more than 5% change in the parameter estimate of any exposure category.

Multiple switching
Analysing only first switches would neglect a portion of the costs generated by patients who switched NSAIDs more than once during follow-up. In a separate analysis that considered all switches during follow-up, a given patient could contribute person-days of NSAID exposure to more than one of the four drug categories. Hazard ratios for switching were estimated using a Cox regression model for multiple events that adjusted for number of previous switches, baseline patient characteristics and treatment duration. Independent variable inclusion criteria were the same as for the first-switch Cox model.

Costs of switching-related healthcare resources
Total costs (mean ± S.D.) of switching-related healthcare resources were calculated per switch and per prescription. Because cost data were skewed and most patients had no switching-related costs, a three-step estimation procedure was used to compare the direct costs of switching-related healthcare resources associated with switching for rofecoxib, celecoxib and Arthrotec users versus other nsNSAID users. First, the HR of switching was estimated by the Cox model for multiple events. Second, the ratio of mean switching-related healthcare costs among switchers was estimated for rofecoxib versus nsNSAIDs other than Arthrotec, celecoxib versus nsNSAIDs other than Arthrotec, and Arthrotec versus other nsNSAIDs using a log-linear model that adjusted for independent variables included in the Cox model. Third, the HR was multiplied by the ratio of the estimated mean switching-related healthcare costs among switchers to estimate the ratio of the mean switching-related healthcare costs among all patients. For example, a ratio of less than 1 for rofecoxib versus nsNSAIDs other than Arthrotec would indicate that, averaged over all users, rofecoxib users generated lower costs of switching-related healthcare resources. The 95% confidence intervals (CIs) for these ratios were estimated from the standard deviations of the estimates given by the Cox and log-linear models. Because the imprecision of the estimate yielded by the three-step procedure is highest when the correlation ({rho}) between estimates from the two models is {rho} = 1 and lowest when {rho} = 0, we calculated the 95% confidence intervals under both assumptions.

All statistical analyses were performed using SAS version 8.2 (SAS Institute Inc., Cary, NC, USA).


   Results
Top
 Abstract
 Introduction
 Methods
 Statistical analyses
 Results
References
 
Patient baseline characteristics
Of the 16 881 patients included in the study, 8896 received rofecoxib at index date, 6676 received celecoxib, 261 received Arthrotec and 1048 received other nsNSAIDs. The relatively low proportion of nsNSAID recipients reflects both the advanced age of the study cohort and the unrestricted coverage of coxibs by the Quebec drug plan. Baseline characteristics are shown in Table 1. Compared with recipients of non-Arthrotec nsNSAIDs, patients on coxibs were more likely to be older, be female, have osteoarthritis, ischaemic heart disease or cerebrovascular disease, have been hospitalized for gastroduodenal perforations, ulcers or bleeding (PUB) and have received anticoagulant therapy or a GPA in the year prior to the index date. Patients receiving coxibs at the index date were less likely to also fill a GPA prescription at the index date than patients on nsNSAIDs other than Arthrotec. Among nsNSAID users, patients receiving Arthrotec were more likely to be older, be female and have musculoskeletal disease or cancer than patients receiving other nsNSAIDs. Unsurprisingly, Arthrotec users were less likely than other nsNSAID users to fill a separate GPA prescription at the index date.

Drug exposure and switching
Figure 1 shows that more patients who received an nsNSAID at the index date switched at least once in the follow-up compared with those who received a coxib. The unadjusted first-switch rates were higher in the nsNSAID group than in either coxib group. The percentage of patients who switched at least once and unadjusted first-switch rates were both higher for Arthrotec recipients than for patients who received other nsNSAIDs.


Figure 1
View larger version (13K):
[in this window]
[in a new window]
  		FIG. 1. Unadjusted incidence (bars) and rate (= number of patients who switched medications at least once divided by total number of days until the first switch or the end of follow-up for those who did not switch; symbols) of a first switch during follow-up categorized by drug received at the index date. nsNSAID, non-selective non-steroidal anti-inflammatory drug.

 
Throughout follow-up, patients filled 31 257 prescriptions for celecoxib, 38 239 for rofecoxib and 5496 for nsNSAIDs. Five of the 17 different nsNSAIDs dispensed accounted for 80% of prescriptions (naproxen 23.2%, Arthrotec 21.9%, diclofenac 19.7%, ibuprofen 9.8%, indomethacin 5.0%). Table 2 shows that nsNSAID prescription durations were, on average, shorter than those for either coxib, but this difference was much less pronounced for Arthrotec than for other nsNSAIDs. In addition, the percentages of Arthrotec and other nsNSAID prescriptions interrupted by a switch to a coxib were higher than the percentage of rofecoxib or celecoxib prescriptions interrupted by a switch to the other coxib or an nsNSAID. Approximately two-thirds of switches from a coxib were to the other coxib rather than to an nsNSAID.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Exposure to NSAIDs and switching behaviour (including multiple switching)

 
Adjusted total switching hazard ratios are shown in Table 3. Baseline patient characteristics that were significantly positively associated with switching included corticosteroid use in the prior year, a diagnosed musculoskeletal disease and being treated by a general practitioner versus a specialist at the index date. Among patients with musculoskeletal diseases, those with osteoarthritis were less likely to switch than patients without this diagnosis. Significantly lower hazard ratios for switching were seen for patients older than 84 yr compared with those aged 66–74 yr, patients with concomitant aspirin use compared with those without, and patients who received a GPA at index date compared with those who did not.


View this table:
[in this window]
[in a new window]
  		TABLE 3. Patient characteristics associated with switchinga

 
Compared with users of nsNSAIDs other than Arthrotec, coxib users were at significantly lower risk of switching after adjusting for baseline patient characteristics, whether considering only first switches or all switches during follow-up. Compared with patients receiving nsNSAIDs other than Arthrotec at the index date, those receiving rofecoxib and celecoxib at the index date were 65 and 62%, respectively, less likely to ever switch. Considering all switches, patients were 61 and 57% less likely to switch while exposed to rofecoxib and celecoxib, respectively, than while exposed to nsNSAIDs other than Arthrotec. There was no significant difference in adjusted switching rates between Arthrotec and other nsNSAIDs.

Costs
Unadjusted costs of NSAIDs and switching-related healthcare resources associated with switching throughout follow-up are displayed in Table 4. The mean daily acquisition costs for celecoxib, rofecoxib and Arthrotec prescriptions were higher than for other nsNSAIDs ($1.66, $1.49, $1.44 and $0.86, respectively). Because on a per-switch basis the cost of wasted medications was higher for coxibs than for nsNSAIDs, and the costs of switch-associated physician visits were similar among the four NSAID groups, the overall cost of switching-related healthcare resources was higher for coxibs than nsNSAIDs. The overall cost of switching-related healthcare resources was higher for Arthrotec than other nsNSAIDs. However, on a per prescription basis, the unadjusted costs of switching-related healthcare resources were higher for nsNSAIDs than for either coxib, due to the higher rate of switching during nsNSAID exposure. This cost difference was more pronounced for Arthrotec than for other nsNSAIDs.


View this table:
[in this window]
[in a new window]
  		TABLE 4. Unadjusted costs (Can$) of NSAIDs and switching-related healthcare resources associated with switching (including multiple switching)

 
Table 5 shows the results of the log-linear model (step two in the three-step estimation procedure) comparing switching-related healthcare costs among switchers in the four NSAID groups. Among patients who switched NSAIDs, the average rofecoxib, celecoxib and Arthrotec prescription was associated with 35, 42 and 28% higher adjusted switching-related costs, respectively, than the average non-Arthrotec nsNSAID prescription.


View this table:
[in this window]
[in a new window]
  		TABLE 5. Ratio of mean costs of switching-related healthcare resources among switchers and among all patientsa

 
Table 5 also presents the end result of the three-step estimation procedure, comparing switching-related healthcare costs among all patients exposed to either coxib versus those to nsNSAIDs. For all patients throughout follow-up, compared with non-Arthrotec nsNSAID prescriptions, average switching-related healthcare costs were 34% higher for Arthrotec prescriptions, and 53 and 47% lower for rofecoxib and celecoxib prescriptions, respectively. Results were robust to varying assumptions about the correlation between estimates from the Cox multiple event and log-linear models, with only slight changes in the 95% CIs as {rho} was varied between 0 and 1.

Discussion
Our results demonstrate that patients using nsNSAIDs were more likely to switch to a coxib than patients using either rofecoxib or celecoxib were to switch to the other coxib or an nsNSAID. This may be explained by the higher likelihood of experiencing upper GI symptoms among nsNSAID users than coxib users [5–7, 9]. Prescription of a GPA at the index date decreased the risk of switching, supporting this hypothesis. The possibility of switching for reasons related to response to therapy also exists, but could not be assessed because the RAMQ database contains no direct measure of clinical effectiveness (i.e. pain relief).

The higher likelihood of switching seen among nsNSAID users is consistent with results from clinical trials and retrospective database analyses. Randomized clinical trials have reported significantly higher rates of treatment discontinuation due to lack of tolerability among nsNSAID recipients than among coxib recipients [5–7]. Several retrospective database studies have found greater persistence with treatment and lower switching rates among coxib users than among nsNSAIDs users [24–27].

Unlike the studies by Zhao et al. [25] and Bennett et al. [34], we did not find that patients at higher risk for serious GI adverse events, such as those with prior GI hospitalizations or users of anticoagulants or aspirin, switched more than those who did not. This might be because Quebec patients at higher risk for GI adverse events are more likely to receive a coxib or a GPA co-prescription [31, 35].

Patients with recorded diagnoses of musculoskeletal conditions switched more than those without such diagnoses. This may be due to the experience of more intense pain, possibly resulting in ineffectiveness of prescribed NSAIDs or the need for higher nsNSAID doses that are potentially associated with a higher risk for GI side-effects. Gastrointestinal side-effects may also be responsible for the higher switching rate among patients who had received corticosteroids in the prior year, since corticosteroid use is a known risk factor for NSAID-related gastropathy [1].

Our results showed that the costs of wasted pills and physician visits associated with medication switches were twice as high among nsNSAID recipients as among coxib recipients; these costs were even higher among Arthrotec users. Costs of wasted resources were slightly but not significantly higher for celecoxib users than rofecoxib users. This is perhaps because multiple daily dosing is common among celecoxib recipients, whereas most patients on rofecoxib took one pill per day [36].

The main strength of this analysis is the use of a large population-based claims database, which has been well validated [37]. This database includes patients with multiple GI risk factors, who are typically excluded from clinical trials but whose treatment is highly relevant to clinical practice. Moreover, this study used actual costs reimbursed by RAMQ, reflecting filled prescriptions; thus, switching patterns reflect real-life utilization.

Nevertheless, this study has limitations. First, the drug exposure groups could be subject to selection bias [38] because differences may exist between groups regarding variables unavailable in the database. For example, OTC medication purchases are not available from RAMQ and could not be assessed (ibuprofen is the only nsNSAID available OTC). However, since patients were included in this study cohort only if they had an NSAID prescription, they are less likely to also use OTC nsNSAIDs. Patients had a financial incentive to obtain nsNSAIDs via prescriptions rather than OTC, since only prescriptions are reimbursed by RAMQ.

A related limitation is that coxib recipients tended to be older and in poorer health than nsNSAID recipients and they may have been more likely to have tried nsNSAIDs in the past, either by prescription or OTC. This could inflate the difference in switch rate between coxib and nsNSAID users because prior nsNSAID intolerance would be expected to reduce the likelihood of switching to an nsNSAID during follow-up. This potential bias was partially controlled for by excluding patients who had received NSAID prescriptions in the prior year, but it was not possible to control for prescription NSAID use earlier than 1 yr before the index date or for prior OTC nsNSAID use, as explained above.

Another limitation is that NSAIDs are not always taken as prescribed, and patients who filled prescriptions for these drugs may not have consumed them. However, we included only patients who filled at least two consecutive prescriptions without interruption and these chronic users may be more likely to have consumed the pills prescribed. Restricting analysis to chronic users also increased the homogeneity of the study population; switching among sporadic NSAID users might be less reflective of drug efficacy and tolerability than among chronic users.

If a patient did not refill an NSAID prescription, follow-up was terminated on the first date with no NSAID exposure. However, it was not possible to determine the date when a patient actually stopped NSAID therapy. If patients stopped therapy before the end of intended therapy duration, this would have resulted in additional costs for wasted medication and possibly physician visits if therapy was stopped due to lack of tolerability. Although these costs could not be determined, restricting analysis to chronic NSAID users should have reduced the number of patients who could have stopped NSAID therapy altogether, given their ongoing need for pain relief. Furthermore, some patients with apparent stoppages in NSAID use probably resumed taking NSAIDs later, before filling a new prescription. Such patients may be expected to consume pills remaining from their last prescription, without wastage.

With the exception of Arthrotec, we did not account for switches from one nsNSAID to another and their associated costs. Therefore, the true cost of switching-related healthcare resources is undoubtedly higher among nsNSAID recipients than estimated in this study. Including all 17 nsNSAIDs prescribed during follow-up would have complicated the analyses. Sample sizes for some nsNSAIDs were too small to yield robust model results, and some nsNSAIDs were never taken at the index date, only after switching.

Finally, all diagnoses in the RAMQ database are based on physician billings and are subject to misclassification. A recent comparison of ICD-9 codes recorded in the RAMQ database with physician charts found that the RAMQ database has limited sensitivity for certain medical conditions [39], potentially reducing the power of statistical tests. Some unknown fraction of the patients with no diagnosed musculoskeletal conditions may have received NSAIDs for musculoskeletal pain, even though the diagnosis was not recorded, because the primary reason for the visit was another medical complaint. This would reduce the ability of the Cox regression models to detect differences in switching rates between patients with and without musculoskeletal conditions. It is probable, however, that patients with unrecorded musculoskeletal conditions had milder symptoms than those diagnosed with these conditions.

In summary, our results suggest that among new users of NSAIDs, after adjusting for confounding factors, patients taking a coxib are less likely to switch therapy than those receiving nsNSAIDs. The adjusted daily cost of switching-related wasted resources was approximately twice as high for nsNSAID users compared with users of either rofecoxib or celecoxib. Prescribing patterns for NSAIDs in Quebec have changed since this study was conducted, with the withdrawal of rofecoxib in September 2004. Approximately 17% of patients in our study had a history of ischaemic heart disease, which may be listed as a contraindication in coxib and nsNSAID labels in the future [14, 15]. Nevertheless, these results remain important and applicable with regard to today's clinical practice, as switching of NSAID therapy occurs frequently and the respective costs are often not considered when making therapy decisions. Physicians and policy makers should take into consideration the likelihood of switching and associated wasted resources and additional healthcare costs generated by a suboptimal therapy.
Figure 2


   Acknowledgments
 
This study was supported by a grant from Merck & Co., Inc. This article was prepared with the assistance of BioMedCom Consultants Inc., Montreal, Canada.

Elham Rahme has received grants from Merck & Co., Inc., the manufacturer of rofecoxib, and from Pfizer, Inc., the manufacturer of celecoxib. She has also served as a consultant for both companies. Elke Hunsche is currently employed by Merck & Co., Inc. Youssef Toubouti reports no potential conflict of interest. On 30 September 2004, Merck & Co., Inc., announced the voluntary worldwide withdrawal of rofecoxib from the market.


   References
Top
 Abstract
 Introduction
 Methods
 Statistical analyses
 Results
 References
 

  1. Laine L. Approaches to nonsteroidal anti-inflammatory drug use in the high-risk patient. Gastroenterology 2001;120:594–606.[CrossRef][Web of Science][Medline]
  2. Talley NJ, Evans JM, Fleming KC, Harmsen WS, Zinsmeister AR, Melton LJ III. Nonsteroidal antiinflammatory drugs and dyspepsia in the elderly. Dig Dis Sci 1995;40:1345–50.[CrossRef][Web of Science][Medline]
  3. Griffin MR. Epidemiology of nonsteroidal anti-inflammatory drug-associated gastrointestinal injury. Am J Med 1998;104: 23S–29S.[Medline]
  4. Hernandez-Diaz S, Rodriguez LA. Association between nonsteroidal anti-inflammatory drugs and upper gastrointestinal tract bleeding/perforation: an overview of epidemiologic studies published in the 1990s. Arch Intern Med 2000;160:2093–9.[Abstract/Free Full Text]
  5. Lisse JR, Perlman M, Johansson G et al. Gastrointestinal tolerability and effectiveness of rofecoxib versus naproxen in the treatment of osteoarthritis: a randomized, controlled trial. Ann Intern Med 2003;139:539–46.[Abstract/Free Full Text]
  6. Bombardier C, Laine L, Reicin A et al. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group. N Engl J Med 2000;343:1520–8.[Abstract/Free Full Text]
  7. Silverstein FE, Faich G, Goldstein JL et al. Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: a randomized controlled trial. Celecoxib Long-term Arthritis Safety Study. J Am Med Assoc 2000;284:1247–55.[Abstract/Free Full Text]
  8. Laine L, Connors LG, Reicin A et al. Serious lower gastrointestinal clinical events with nonselective NSAID or coxib use. Gastroenterology 2003;124:288–92.[CrossRef][Web of Science][Medline]
  9. Deeks JJ, Smith LA, Bradley MD. Efficacy, tolerability, and upper gastrointestinal safety of celecoxib for treatment of osteoarthritis and rheumatoid arthritis: systematic review of randomised controlled trials. Br Med J 2002;325: 619.[Abstract/Free Full Text]
  10. Jüni P, Rutjes AW, Dieppe PA. Are selective COX 2 inhibitors superior to traditional non steroidal anti-inflammatory drugs? Br Med J 2002;324:1287–8.[Free Full Text]
  11. Aw TJ, Haas SJ, Liew D, Krum H. Meta-analysis of cyclooxygenase-2 inhibitors and their effects on blood pressure. Arch Intern Med 2005;165:490–6.[Abstract/Free Full Text]
  12. Bresalier RS, Sandler RS, Quan H et al. Cardiovascular events associated with rofecoxib in a colorectal adenoma chemoprevention trial. N Engl J Med 2005;352:1092–102.[Abstract/Free Full Text]
  13. Hampton T. Officials halt NSAID prevention trials. J Am Med Assoc 2005;293:664–5.[Free Full Text]
  14. Health Canada. Report of the Expert Advisory Panel on the safety of Cox-2 selective non-steroidal anti-inflammatory drugs (NSAIDs). Health Canada website (accessed 19 August 2005). (http://www.hc-sc.gc.ca/dhp-mps/prodpharma/activit/sci-consult/cox2/sap_report_gcs_rapport_cox2_e.html).
  15. Jenkins JJ, Seligman PJ (US Food and Drug Administration). Memorandum: analysis and recommendations for Agency action regarding non-steroidal anti-inflammatory drugs and cardiovascular risk. Center for Drug Evaluation and Research website (accessed 19 August 2005). (http://www.fda.gov/cder/drug/infopage/COX2/NSAIDdecisionMemo.pdf).
  16. Lévesque LE, Brophy JM, Zhang B. The risk for myocardial infarction with cyclooxygenase-2 inhibitors: a population study of elderly adults. Ann Intern Med 2005;142:481–9.[Abstract/Free Full Text]
  17. Solomon SD, McMurray JJ, Pfeffer MA et al. Cardiovascular risk associated with celecoxib in a clinical trial for colorectal adenoma prevention. N Engl J Med 2005;352:1071–80.[Abstract/Free Full Text]
  18. Sowers JR, White WB, Pitt B et al. The effects of cyclooxygenase-2 inhibitors and nonsteroidal anti-inflammatory therapy on 24-hour blood pressure in patients with hypertension, osteoarthritis, and type 2 diabetes mellitus. Arch Intern Med 2005;165:161–8.[Abstract/Free Full Text]
  19. Bleumink GS, Feenstra J, Sturkenboom MC, Stricker BH. Nonsteroidal anti-inflammatory drugs and heart failure. Drugs 2003;63:525–34.[CrossRef][Web of Science][Medline]
  20. Ray WA, Stein CM, Hall K, Daugherty JR, Griffin MR. Non-steroidal anti-inflammatory drugs and risk of serious coronary heart disease: an observational cohort study. Lancet 2002;359:118–23.[CrossRef][Web of Science][Medline]
  21. Langman M, Kahler KH, Kong SX et al. Drug switching patterns among patients taking non-steroidal anti-inflammatory drugs: a retrospective cohort study of a general practitioners database in the United Kingdom. Pharmacoepidemiol Drug Saf 2001;10: 517–24.[CrossRef][Medline]
  22. Sift R, van Staa TP, Abenhaim L, Ebner D. A study of the longitudinal utilization and switching-patterns of non-steroidal anti-inflammatory drugs using a pharmacy based approach. Pharmacoepidemiol Drug Saf 1997;6:263–8.[Medline]
  23. Walker AM, Chan KW, Yood RA. Patterns of interchange in the dispensing of non-steroidal anti-inflammatory drugs. J Clin Epidemiol 1992;45:187–95.[CrossRef][Web of Science][Medline]
  24. Wolfe F, Michaud K, Burke TA, Zhao SZ. Longer use of COX-2-specific inhibitors compared to nonspecific nonsteroidal antiinflammatory drugs: a longitudinal study of 3639 patients in community practice. J Rheumatol 2004;31:355–8.[Abstract/Free Full Text]
  25. Zhao SZ, Wentworth C, Burke TA, Makuch RW. Drug switching patterns among patients with rheumatoid arthritis and osteoarthritis using COX-2 specific inhibitors and non-specific NSAIDs. Pharmacoepidemiol Drug Saf 2004;13:277–87.[CrossRef][Web of Science][Medline]
  26. Harley C, Wagner S. Persistence with COX-2 inhibitors in managed care: an analysis of claims data. Manag Care Interface 2003;16:38–45.[Medline]
  27. Moride Y, Ducruet T, Rochon S, Lavoie F. Persistency of use of COX-2-specific inhibitors and non-specific non-steroidal anti-inflammatory drugs (NSAIDs) in Quebec. Rheumatology 2003;42(Suppl 3): iii17–iii22.
  28. Rahme E, Joseph L, Kong SX, Watson DJ, LeLorier J. Gastrointestinal health care resource use and costs associated with nonsteroidal antiinflammatory drugs versus acetaminophen: retrospective cohort study of an elderly population. Arthritis Rheum 2000;43:917–24.[CrossRef][Web of Science][Medline]
  29. World Health Organization. Adherence to long-term therapies: evidence for action. Geneva, World Health Organization, 2003.
  30. ICD9.CM millennium edition: International Classification of Diseases, ninth revision, clinical modification, 6th edn. Los Angeles: Practice Management Information Corporation (PMIC), 2003.
  31. Rahme E, Marentette MA, Kong SX, LeLorier J. Use of NSAIDs, COX-2 inhibitors, and acetaminophen and associated coprescriptions of gastroprotective agents in an elderly population. Arthritis Rheum 2002;47:595–602.[CrossRef][Web of Science][Medline]
  32. Hooper L, Brown TJ, Elliott R, Payne K, Roberts C, Symmons D. The effectiveness of five strategies for the prevention of gastrointestinal toxicity induced by non-steroidal anti-inflammatory drugs: systematic review. Br Med J 2004;329:948–57.[Abstract/Free Full Text]
  33. Federal Reserve Statistical Release. G.5 A: foreign exchange rates (annual). US Federal Reserve Board website (accessed 11 March 2005) (http://www.federalreserve.gov/releases/g5a/20030106/).
  34. Bennett K, Teeling M, Feely J. ‘Selective’ switching from non-selective to selective non-steroidal anti-inflammatory drugs. Eur J Clin Pharmacol 2003;59:645–9.[CrossRef][Web of Science][Medline]
  35. Moride Y, Ducruet T, Boivin JF, Moore N, Perreault S, Zhao S. Prescription channeling of COX-2 inhibitors and traditional nonselective nonsteroidal anti-inflammatory drugs: a population-based case-control study. Arthritis Res Ther 2005;7: R333–R342.[Medline]
  36. Schnitzer TJ, Kong SX, Mitchell JH et al. An observational, retrospective, cohort study of dosing patterns for rofecoxib and celecoxib in the treatment of arthritis. Clin Ther 2003;25:3162–72.[Medline]
  37. Tamblyn R, Lavoie G, Petrella L, Monette J. The use of prescription claims databases in pharmacoepidemiological research: the accuracy and comprehensiveness of the prescription claims database in Quebec. J Clin Epidemiol 1995;48:999–1009.[CrossRef][Web of Science][Medline]
  38. Savitz A. Interpreting epidemiologic evidence: strategies for study design and analysis. New York, NY: Oxford University Press, 2003.
  39. Wilchesky M, Tamblyn RM, Huang A. Validation of diagnostic codes within medical services claims. J Clin Epidemiol 2004;57:131–41.[CrossRef][Web of Science][Medline]
Submitted 7 July 2005; revised version accepted 22 November 2005.
Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
45/7/903    most recent
kei269v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (3)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Rahme, E.
Right arrow Articles by Hunsche, E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rahme, E.
Right arrow Articles by Hunsche, E.
Related Collections
Right arrow Health Economics
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?