Embed Size (px)
Citation preview
Health Policy 84 (2007) 1–13
The effects of coxib formulary restrictions on analgesicuse and cost: Regional evidence from Canada
Deborah A. Marshall a,b,∗, Donald J. Willison a,b, Paul Grootendorst c,d,Jacques LeLorier e,f, Malcolm Maclure g,h, Nathalie A. Kulin a, Odile E. Sheehy f,
Leanne Warren h, Kathy Sykora i, Elham Rahme j
a Centre for Evaluation of Medicines, 105 Main Street East, Level P1, Hamilton, ON L8N 1G6, Canadab Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, ON, Canada
c Faculty of Pharmacy, University of Toronto, Toronto, ON, Canadad Department of Economics, McMaster University, Hamilton, ON, Canada
e Departments of Medicine and Pharmacology, Faculty of Medicine, University of Montreal, Montreal, QC, Canadaf Pharmacoepidemiology and Pharmacoeconomics Research Unit, Centre hospitalier de l’Universite de Montreal,
Montreal, QC, Canadag School of Health Information Science, University of Victoria, BC, Canada
h Pharmaceutical Services Division, Ministry of Health, BC, Canadai Institute for Clinical Evaluative Sciences, Toronto, ON, Canada
j Department of Medicine, Division of Clinical Epidemiology, Research Institute of the McGillUniversity Health Centre, Montreal, QC, Canada
Abstract
Background: Public insurance plans for pharmaceuticals in Canada differ substantially across provinces in the conditions underwhich pharmaceuticals are reimbursed. Coxibs provide a good example. Quebec had no restrictions on reimbursement for thesedrugs. Ontario required physicians to submit the clinical indications for their use on the prescription. British Columbia requiredphysicians to seek and receive prior authorisation from the drug plan.Objective: This study compares the effects of different reimbursement policies on coxib, non-selective non-steroidal anti-inflammatory drugs (nsNSAIDs), and gastro-protective agent (GPA) use and cost.Study design: Analysis of retrospective time series analysis of all NSAID and GPA administrative claims data from April 1997through December 2002.
Setting: Administrative claims data from April 1997 through December 2002 for each of the publicly funded drug plansin Quebec, Ontario, and British Columbia. In addition, we obtained data from BC PharmaNet, which records all dispensedprescriptions in British Columbia.Patients or other participants: Senior beneficiaries (≥ 65 years).∗ Corresponding author at: Centre for Evaluation of Medicines, 105 Main Street East, Level P1, Hamilton, ON L8N 1G6, Canada.Tel.: +1 905 522 1155x34967; fax: +1 905 528 7386.
E-mail address: [email protected] (D.A. Marshall).
0168-8510/$ – see front matter © 2007 Elsevier Ireland Ltd. All rights reserved.doi:10.1016/j.healthpol.2007.04.010
2
MwnRta(wCo©
Ka
1
cefadto
ipTmmbwbdpppeda
btuclA
D.A. Marshall et al. / Health Policy 84 (2007) 1–13
ain outcome measure: We compared the projected total NSAID utilisation in the absence of coxib reimbursement restrictionith actual utilisation by province and drug category. Projected utilisation was based on ARIMA modelling and reported as theumber of defined daily doses (DDDs) per 100 senior (≥65 years) beneficiaries/month.esults: In Ontario, under its “limited use” policy, uptake and steady-state use of coxibs was similar to that in Quebec, where
here were no restrictions. In British Columbia, publicly funded use of coxibs was 6% of that in Ontario and Quebec. Despiteshift to private reimbursement, total coxib use in BC was only 50% of use in Ontario and Quebec. The use of all NSAIDS
nsNSAIDS plus coxibs) increased for all provincial drug plans except for BC. The increase and overall rate of total NSAID useas greatest in Ontario. Neither Ontario’s nor BC’s policies had an impact on use of nsNSAIDs or GPAs.onclusion: Only BC’s policy effectively limited publicly funded coxib use. However, there was substantial cost-shifting tout-of-pocket and third party insurance plans in BC.
2007 Elsevier Ireland Ltd. All rights reserved.
hibitors
otn(mmugetcosstr
tQprBjtiv“t(
eywords: Non-steroidal anti-inflammatory drugs; Coxibs; Cox-2 inuthority; Limited use policy; Analgesics
. Background
Globally, growth in expenditures for pharmaceuti-als has outpaced the average growth in health carexpenditures. In Canada, pharmaceuticals have risenrom 9.5% of expenditures in 1985 to 16.1% in 2002nd were estimated to reach 16.7% in 2004 [1]. Expen-itures on drugs in Canada are now surpassed only byhat spent on hospitals, driven largely by the adoptionf newer drugs [2,3].
About half of prescription drug spending in Canadas publicly financed, with provincial government druglans accounting for the majority of public spending.here is no common standard for coverage for phar-aceuticals across provinces. Each provincial govern-ent determines independently which drugs are reim-
ursed, beneficiary co-payments, and the criteria underhich drugs are reimbursed. Faced with constrainedudgets, provincial policy makers have independentlyeveloped and implemented various administrativeolicies to restrict the use of certain pharmaceuticals toatients who meet specific clinical criteria. This inter-rovincial variation in drug policy has created naturalxperiments with which one can compare the impact ofifferent types of reimbursement policies on the uptakend use of new and existing drugs.
Cyclooxenase-2 inhibitors (coxibs) exemplify welloth the challenges faced by drug plans and the impacthat alternative reimbursement policies have on the
ptake of new pharmaceuticals. Celecoxib was the firstoxib approved by Health Canada in April 1999, fol-owed by rofecoxib in October 1999, and meloxicam inugust 2000. On introduction, the purported advantagettpu
; Health policy; Drug diffusion; Administrative restrictions; Special
f the coxibs was a lower incidence of severe gastroin-estinal events (bleeding and ulcers), as compared withon-selective non-steroidal anti-inflammatory drugsnsNSAIDs) [4–6]. Coxibs cost substantially more thanany conventional NSAIDs, and well over 10 timesore than acetaminophen or ibuprofen. Coxib man-
facturers justified their relatively high price on therounds that coxibs would reduce the use of – andxpenditures on – gastro-protective agents, which areypically taken concomitantly with nsNSAIDs. Rofe-oxib was withdrawn from the market in 2005 becausef increased incidence of myocardial infarction andtroke [7–12]. Therefore, there are both economic andafety reasons for ensuring that coxibs are used only inhose patients for whom there is a favorable benefit–riskatio.
Reimbursement of coxibs varies markedly amonghe provincial government drug plans in Canada. Inuebec, coxibs are listed as a general benefit on therovincial formulary. They can be prescribed with noestrictions on reimbursement. Under the Ontario Drugenefits (ODB) Program, physicians are required to
ustify a coxib at the time of prescribing by providinghe appropriate code on a special prescription form,ndicating why a coxib is required instead of a con-entional nsNSAID. This approach is referred to asLimited Use” and applies to some 150 drug enti-ies in Ontario [13]. In contrast, the British ColumbiaBC) PharmaCare program requires prior authorisa-
ion before it reimburses prescriptions for coxibs. Inhis paper, we examine how the coxib reimbursementolicies in BC (prior authorisation), Ontario (limitedse), and Quebec (no restrictions) have affected the
ealth P
pttIicdadi
2
2
bfdabboPitdpbaurdmb
2
oSdgo(pd
chplt
2
2
fHtrwura
2
OpHcOtupaOppaictosubleed, OR treatment failure or intolerance to at leastthree listed nsNSAIDS. A trial of acetaminophen isnot required for patients with rheumatoid arthritis. The
D.A. Marshall et al. / H
rescribing of coxibs, their anti-inflammatory substi-utes (nsNSAIDs, diclofenac sodium/misoprostol), andheir complements, gastro-protective agents (GPAs).n particular, we ask: how do these different prescrib-ng restrictions affect the uptake and expenditures onoxibs? Is Ontario’s Limited Use policy an effectiveeterrent to the prescribing of coxibs? To what extentre increased coxib expenditures in jurisdictions whicho not restrict their use (Quebec) offset by reductionsn GPA expenditure?
. Methods
.1. Data sources
Aggregate data on anti-inflammatory and GPA useetween April 1997 and December 2002 were obtainedrom each of the three provincial drug plans (Regiee l’assurance maladie du Quebec (RAMQ), ODB,nd BC PharmaCare). These data reported the num-er of prescriptions filled and the total reimbursementy the drug plan on a monthly basis. In addition, webtained data on total coxib use in BC through the BCharmaNet, which records all prescriptions dispensed
n community pharmacies in the province, irrespec-ive of who paid for the prescription. The PharmaNetata, therefore, consist of dispensings of medicationsaid by PharmaCare plus those paid out-of-pocket ory federal government or private third-party insur-nce. For simplicity, we refer to prescription drugse covered by PharmaCare through its prior autho-isation process as “PharmaCare” and prescriptionrug use not covered by PharmaCare as “Phar-aNet”, although the PharmaNet database includes
oth.
.2. Study population
We focused on senior beneficiaries (65 years orlder) between April 1, 1997, and December 31, 2002.eniors are eligible for coverage from their provincialrug plan and also have particularly high rates of anal-esic use. For the analyses of GPA diffusion, instead
f all seniors, the denominator was all “at-risk” seniorsi.e., the number of seniors who received at least onerescription of an nsNSAID, coxib, or acetaminophenuring the study period).Id
olicy 84 (2007) 1–13 3
Coxibs consisted of celecoxib, rofecoxib, meloxi-am, and valdecoxib.1 Included in the GPA group wereistamine H2-receptor antagonists (e.g., ranitidine) androton pump inhibitors (e.g., omeprazole). A completeist of nsNSAIDs and GPAs analysed is available fromhe authors.
.3. Provincial drug plans
.3.1. QuebecCelecoxib and rofecoxib were added to the RAMQ
ormulary six months after they were approved byealth Canada in October 1999 and April 2000, respec-
ively. Because most coxibs (including celecoxib,ofecoxib, meloxicam, and etodolac) were reimbursedithout prescribing restrictions, Quebec serves as aseful control to compare the impact of the moreestrictive reimbursement policies adopted in Ontariond BC.
.3.2. OntarioCelecoxib and rofecoxib were introduced to the
DB formulary in April 2000 as “Limited Use” (LU)roducts, 1 year after celecoxib was approved byealth Canada. Limited use products are those not
onsidered appropriate for a general listing on theDB formulary, but may be reimbursed when cer-
ain clinical criteria are met. Coxibs were deemednsuitable as first line NSAID therapy for arthriticatients since they were costlier than alternatives avail-ble through the ODB general benefit listing. Anntario physician wishing to write an ODB-eligiblerescription for an LU product must ensure that theatient meets the clinical criteria outlined by ODBnd provide the correct code on the prescription form,ndicating that the clinical criteria for prescribing aoxib have been met. For osteoarthritis, these are: (a)reatment failure or intolerance to an adequate trialf acetaminophen (e.g., acetaminophen 1 g QID foreveral weeks), and have had (b) a history of a doc-mented, clinically significant ulcer or gastrointestinal
1 Valdecoxib was approved by Health Canada in December 2002.t was not included on any of the provincial formularies of interesturing the study time horizon.
4 ealth P
mbabbtjwMfM
2
PiHrPfdtffttntorbActb
2
uphtpvtTm
edi1nca
obsbroabecsib
pcp
pdamDb
ttmmitdctt
D.A. Marshall et al. / H
aximum daily dose of coxib which will be reim-ursed is specified for the treatment of osteoarthritisnd rheumatoid arthritis. The prescription form muste rewritten annually. There is no administrative delayefore the patient may receive the medication but bothhe pharmacist and the prescribing physician are sub-ect to audit by the Ontario government. As far ase are aware, however, no audits have ever occurred.eloxicam was subsequently introduced to the Ontario
ormulary as a general (i.e., unrestricted) benefit inarch 2001.
.3.3. British ColumbiaCelecoxib and rofecoxib were listed on the BC
harmaCare formulary in September 2000, approx-mately 1.5 years after celecoxib was approved byealth Canada. PharmaCare has the most stringent
eimbursement restrictions of the three provinces. BCharmaCare must approve a Special Authority requestrom the prescribing physician prior to the drug beingispensed, otherwise PharmaCare will not reimbursehe patient. Physicians must document on a faxedorm that beneficiaries meet all the following criteriaor osteoarthritis: (a) treatment failure or intoleranceo acetaminophen; (b) treatment failure or intoleranceo at least one of enteric-coated ASA, ibuprofen, oraproxen; and (c) treatment failure or intoleranceo at least three other specified NSAIDS. A trialf acetaminophen is not required for patients withheumatoid arthritis or other inflammatory conditions,ut criteria (b) and (c) above would still need to be met.pproval of a Special Authority request for a coxib
an take up to 10 business days but is usually lesshan a week [14]. Urgent requests are filled within 1usiness day.
.4. Analytical methods
This study was a retrospective time series analysissing administrative databases from the public druglans in each of the three provinces. In BC, we alsoad access to total population drug utilisation datahrough BC PharmaNet. Because prescribing patterns,rescription size, and the strength of drugs dispensed
aries by province, prescription claims were convertedo the number of defined daily doses (DDDs) dispensed.he DDD is the total quantity of drug dispensed inilligrams divided by the World Health Organisationmmrw
olicy 84 (2007) 1–13
stimates of the typical maintenance dose (mg used peray) [15]. The typical daily doses for the drugs analysedncluded 200 mg for celecoxib, 12.5 mg for rofecoxib,200 mg for ibuprofen, and 500 mg for naproxen. Theumber of DDDs per month was grouped into five drugategories: coxibs; nsNSAIDS; acetaminophen; GPAs;nd total NSAIDs (coxibs plus nsNSAIDs).
To account for the variation in dosing and frequencyf prescription and the different size of the drug planeneficiary population in each province, DDDs weretandardised as the number of DDDs per 100 senioreneficiaries per month [16]. For example, a monthlyate of 300 DDDs per 100 seniors would occur if 10%f seniors took the defined dose of the drug each day inparticular month. The number of provincial drug planeneficiaries 65 years and older in each province wasstimated from census count data for each province,onducted every 5 years [17]. Estimates between cen-us years were interpolated using a linear assumptionn order to estimate the senior beneficiary populationy month.
The month when coxibs were first listed on therovincial formulary was used to demarcate the pre-oxib and post-coxib periods for each analysis byrovince (Table 1).
Diffusion curves were plotted to examine actualatterns of drug use over time in each province byrug category. We performed time series analyses withutoregressive integrated moving average (ARIMA)odels [18–21] to estimate the projected number ofDDs with 95% confidence intervals in each provincey drug category.
We developed a linear regression model to examinehe effect of different formulary reimbursement restric-ions for coxibs, measured as the difference in average
onthly number of DDDs, relative to Quebec. Theodel allowed for the coxib listing to change both the
ntercept and slope of the pre-formulary listing linearime trend in the outcome variable; seasonal variation inrug use was modelled through the use of monthly indi-ator variables. Standard errors were estimated usinghe Newey–West heteroskedasticity and autocorrela-ion consistent covariance matrix estimator [22].
Time series analysis using exponential smoothing
odels and autoregressive integrated moving averageodels were conducted using SAS Version 8.2 andegression analyses using STATA Version 9.1. Graphsere created using Microsoft Excel 2000.
D.A. Marshall et al. / Health Policy 84 (2007) 1–13 5
Table 1Analysis time period definitions of coxib formulary inclusion by jurisdiction
Provincial drug plan Pre-coxib time period Post-coxib time period
Quebec RAMQ April 1997 to September 1999 October 1999 to November 2002a
Ontario Drug Benefit April 1997 to March 2000 April 2000 to December 2002BC PharmaCare April 1997 to September 2000 October 2000 to December 2002BC PharmaNetb,c April 1997 to March 1999 April 1999 to December 2002
BC, British Columbia; coxib, cyclooxenase-2 inhibitor; RAMQ, Regie de l’assurance maladie du Quebec.a The data for December 2002 were incomplete, so December 2002 was excluded from the analysis.
te of ma2
3
3
of
3
a
Fys
b Because BC PharmaNet data capture all drug use, March 1999 (da000 was chosen as the cut-point.c Not covered by BC PharmaCare.
. Results
.1. NSAID utilisation
Figs. 1–3 summarise the trends for the prescribingf coxibs, nsNSAIDs, and total NSAIDs respectively,or each provincial drug plan.
rp1t
ig. 1. Monthly utilisation rates of coxibs by provincial drug plan, betweeears and older). Coxibs included selective non-steroidal anti-inflammatorolid form.
rket introduction in Canada of the first coxib), rather than September
.1.1. Coxib utilisation (Fig. 1)Following their listing in Quebec (September 1999)
nd Ontario (April 2000), rates of coxib use increased
apidly. The use of coxibs rose to a peak of 328 DDDser 100 senior beneficiaries in October 2000 in Quebec,year after their introduction. The initial diffusion pat-ern was similar in Ontario, but occurred 6 months later
n April 1997 and December 2002 (DDDs per 100 beneficiaries 65y drugs (celecoxib, meloxicam, rofecoxib, and valdecoxib) in oral
6 D.A. Marshall et al. / Health Policy 84 (2007) 1–13
F lan, bet6
wcadttiapulaQ
3
tair
toD32fae
3
iBeet
ig. 2. Monthly utilisation rates of nsNSAIDs by provincial drug p5 years and older).
hen coxibs were listed in that province. By June 2001,oxib use in Ontario and Quebec was similar. There-fter, use reached a steady state in Ontario while iteclined in Quebec. Under the BC PharmaCare plan,he uptake of coxibs was much slower and lower, risingo a maximum of 19 DDDs per 100 senior beneficiariesn December 2001—only 6% of that found in Quebecnd Ontario. The BC PharmaNet data (reflecting totalrescribing in the province) indicate that the rate ofptake of all coxibs was slower, and the steady stateevel of use in the entire population of seniors waspproximately 50% of that observed in Ontario anduebec.
.1.2. Non-selective NSAID utilisation (Fig. 2)The use of nsNSAIDs was steadily decreasing over
he study period in all provincial drug plans, andppeared to be unaffected by the introduction of cox-bs. The use of nsNSAIDs per 100 senior beneficiariesemained highest in Ontario and lowest in Quebec
pTas
ween April 1997 and December 2002 (DDDs per 100 beneficiaries
hroughout the time period. The largest decrease wasbserved for RAMQ in Quebec, declining from 148DDs per 100 senior beneficiaries in April 1997 to6 DDDs per 100 senior beneficiaries by November002. The pattern of nsNSAID use was almost identicalor both BC PharmaCare and BC PharmaNet, startingt about 180 DDDs per 100 senior beneficiaries andnding at about 120 DDDs per 100 senior beneficiaries.
.1.3. Total NSAID utilisation (Fig. 3 and Table 2)The use of all NSAIDs (nsNSAIDs plus coxibs)
ncreased for all of the provincial drug plans exceptC PharmaCare (Fig. 3). NSAID use remained high-st in Ontario throughout the time period, with thexception of a transition period at the end of 1999 andhe beginning of 2000, when Quebec’s consumption
er 100 senior beneficiaries exceeded that of Ontario.he absolute increase in the total use of NSAIDs waslso greatest in Ontario – from 213 DDDs per 100enior beneficiaries in April 1997 to 419 DDDs per 100
D.A. Marshall et al. / Health Policy 84 (2007) 1–13 7
F plan, be6
siupa
g
TD
P
QOBB
Yi
2
ig. 3. Monthly utilisation rates of total NSAIDs by provincial drug5 years and older).
enior beneficiaries by April 2002 – 2 years after cox-
bs were listed in Ontario (Table 2). The expansion inse of total NSAIDs in Quebec was initially very rapid,eaking in October 2000, and declining steadily there-fter. By contrast, total NSAID use in BC (PharmaNet)QbtP
able 2iffusion by time since coxibs introduced to market by plan
rovincial drug plan Date of coxiblisting
Utilisation of coxibsa
Year 1 Year 2
uebec RAMQ October 1999 122 328ntario Drug Benefit April 2000 56 270C PharmaCare October 2000b 2 16C PharmaNetc,d April 1999 3 87
ear 1 means 1 year after date of coxib listing; Year 2 means 2 years afternhibitor; DDD, defined daily dose; NSAID, non-steroidal anti-inflammator
a DDD per 100 senior beneficiaries.b List date was 18-Sep-00, but October was used as the cut-off for analysic Because BC PharmaNet data capture all drug use, March 1999 (date of ma000 was chosen as the cut-point.d Not covered by BC PharmaCare.
tween April 1997 and December 2002 (DDDs per 100 beneficiaries
rew more slowly but was virtually identical to that in
uebec at approximately 270 DDDs per 100 senioreneficiaries for both in November 2002. At the end ofhe observation period, NSAID use for RAMQ and BCharmaNet was 68% of that observed in Ontario.Total NSAID utilisation(total market)
% of NSAID market(coxibs as % of total market)
Year 1 Year 2 Year 1 Year 2
218 380 56 86237 419 24 64142 149 2 11167 220 2 39
date of coxib listing. BC, British Columbia; coxib, cyclooxenase-2y drug; RAMQ, Regie de l’assurance maladie du Quebec.
s.rket introduction in Canada of the first coxib), rather than September
8 D.A. Marshall et al. / Health Policy 84 (2007) 1–13
F ween Aa
oaslg1Nwb
3
hvi3nw
p1Bssi
3
oiieaN
ig. 4. Monthly utilisation rates of GPA by provincial drug plan, betnd older).
At their peak, coxibs accounted for 89, 69, and 12%f all NSAIDs publicly reimbursed in Quebec, Ontario,nd BC, respectively (as measured by DDDs per 100enior beneficiaries). Two years after the date of coxibisting in the respective provinces, coxib utilisation wasreatest in Quebec in absolute numbers (328 DDDs per00 senior beneficiaries) and as a proportion of totalSAID utilisation (86%), but total NSAID utilisationas still largest in Ontario (419 DDDs per 100 senioreneficiaries) (Table 2).
.2. GPA utilisation (Fig. 4)
The use of GPAs increased steadily in all provincialealth care plans throughout the window of obser-ation (Fig. 4). The use of GPAs remained highest
n Ontario throughout the time period with 260 to90 DDDs per 100 senior beneficiaries at the begin-ing/end of the observation period, respectively. Thisas followed by Quebec with 120 to 334 DDDscaiP
pril 1997 and December 2002 (DDDs per 100 beneficiaries 65 years
er 100 senior beneficiaries, BC PharmaNet with60 to 243 DDDs per 100 senior beneficiaries andC PharmaCare with 140 to 167 DDDs per 100
enior beneficiaries. Again, there was no observablehift in the slope or intercept associated with coxibntroduction.
.3. Total NSAIDs (Fig. 5 and Table 3)
The effect of administrative restrictions for coxibsn the total NSAID market (nsNSAIDS plus coxibs)s illustrated in Fig. 5, where the pre-formulary list-ng trend in nsNSAIDs use in each jurisdiction isxtrapolated into the post-listing period. There was
statistically significant increase in the use of allSAIDs in the post-coxib period compared to the pre-
oxib period in the Quebec and Ontario drug plansnd BC PharmaNet. There was no statistically signif-cant change in NSAID use coinciding with the BCharmaCare prior authorisation program.
D.A. Marshall et al. / Health Policy 84 (2007) 1–13 9
NSAID
tatpTtoCD
G(1
Oi
TEp
QOBB
Bd
Fig. 5. Actual total NSAID DDDs and projected number of ns
The regression models suggest that the implemen-ation of administrative restrictions for coxibs wasssociated with a statistically significant increase inhe number of DDDs of all NSAIDs for all druglans except the BC PharmaCare drug plan (Table 3).he estimated effect of the introduction of coxibs
o the formulary was greatest in Quebec (increasef 227 DDDs per 100 senior beneficiaries; 95%I: 198–257) and smallest for BC PharmaNet (84DDs per 100 senior beneficiaries; 95% CI: 65–103).
l1P(
able 3stimated average monthly change in total NSAIDs DDD per 100 beneficiarovince and drug
DDDs per 100 senior beneficiaries (9
Average monthly change
uebec RAMQ 227 (198–257, p < 0.001)ntario ODB 200 (180–220, p < 0.001)C PharmaCare 7 (−1 to 15, p = 0.09)C PharmaNeta 84 (65–103, p < 0.001)
C, British Columbia; coxib, cyclooxenase-2 inhibitor; DDD, defined dailye l’assurance maladie du Quebec.a Not covered by BC PharmaCare.
DDDs, time series ARIMA models by provincial drug plan.
rowth in Ontario was slightly less than in Quebec200 DDDs per 100 senior beneficiaries; 95% CI:80–220).
We also estimated the growth in total NSAIDs inntario and BC relative to Quebec (Table 3). Only
n BC was growth in total NSAID use significantly
ower than in Quebec—PharmaCare at−221 DDDs per00 senior beneficiaries (95% CI: −251 to −190) andharmaNet at −144 DDDs per 100 senior beneficiaries95% CI: −178 to −110).ries 65 years and older before and after coxib formulary listing by
5% CI, p-value)
Difference in estimated effects relative to Quebec RAMQ
–−27 (−63 to 8, p = 0.13)−221 (−251 to −190, p < 0.001)−144 (−178 to −110, p < 0.001)
dose; NSAID, non-steroidal anti-inflammatory drug; RAMQ, Regie
1 ealth P
4
cifnt
•
•
•
•
gcibaotfsi
atwft
rcgSdrt
tpipow3slalaeeicai
sIncTracwssi
bces
0 D.A. Marshall et al. / H
. Discussion and conclusions
Our evaluation represents the first multi-provinceomparison of administrative reimbursement policiesn Canada. The variation in drug reimbursement policyound in BC, Ontario, and Quebec provides an idealatural experiment in which to examine the effect ofhese policies. We found that:
By June 2001, coxib use in Ontario and Quebec wasat a similar level (over 300 DDDs per 100 senior ben-eficiaries), despite the limited use restriction policyin Ontario.The delayed formulary listing of the coxibs in BC,coupled with the Special Authority requirement oncethey were listed, limited the use of publicly fundedcoxibs to 6% of the use reported in Quebec andOntario. While some purchased coxibs using privateinsurance or direct payment, total coxib utilisationin BC – both publicly and privately financed – wasabout half of that in Ontario and Quebec.The use of all NSAIDS (nsNSAIDS plus coxibs)increased for all the drug plans except BC Phar-maCare. Total NSAID use was greatest in Ontariocompared to BC and Quebec. Relative to QuebecRAMQ, only in BC was there significantly lowergrowth in total NSAID prescriptions.No corresponding decrease in the use of gastro-protective agents was observed with the increaseduse and substitution of coxibs for nsNSAIDs.
In Canada, an estimated 40% of drug spendingrowth is due to the substitution of high cost for lowost drugs used to treat the same condition [3]. Drugnsurers now routinely require as a condition for reim-ursement that prescribers justify why expensive drugsre required when cheaper alternatives exist. The casef analgesic prescribing for treatment of osteoarthri-is illustrates this phenomenon. In this study, we alsoound not only substitution but, more importantly, sub-tantial expansion in the NSAID market due to coxibsn all markets except BC PharmaCare.
While we cannot state definitively why the BC prioruthorisation policy was effective, it is quite certain
hat the requirement to physically complete and send aritten form signed by the physician for each patientor which a coxib is requested, combined with theime delay in gaining approval, is a significant deter-
Cwho
olicy 84 (2007) 1–13
ent for physicians. Physicians are not reimbursed forompleting the Special Authority request form, and areenerally under severe time constraints. Approval of apecial Authority request can take up to 10 businessays [14]. Thus, an additional follow up visit might beequired to actually provide the written prescription tohe patient after authorisation has been obtained.
One possible additional explanation for the effec-iveness of the BC prior authorisation policy is theublication of the Therapeutics Letter by a BC organ-sation called the Therapeutics Initiative. Althoughrinted materials alone generally have a weak impactn physician behaviours [23], the Therapeutics Letteras found to have shifted prescribing an average of0% from baseline according to an aggregate analy-is of 12 letters using randomised controls [24]. Fouretters were published during the period between 1999nd 2004 [25–28], the first of which pointed to theack of published clinical data on coxib effectivenessnd the other three which focused on serious adversevents associated with coxibs (fewer gastrointestinalvents, but more myocardial infarctions). Otherwise, its our understanding that the policy environment thatould affect utilisation patterns remained stable acrossll three provinces during the time period we are exam-ning.
In each province, nsNSAID utilisation continued itsteady decline in use over the observed time window.t appears that this consistent decrease in the use ofsNSAIDs was independent of the introduction of theoxibs or the patterns of coxib utilisation over time.his general decrease in the use of nsNSAIDs probably
eflects a pre-existing longer-term general movementway from NSAIDs as first line treatment because ofoncerns about the gastrointestinal side effects. Thisould likely be similar across the provinces and, as
uch, this does not impact the overall findings of thistudy, which are focused on the differences in admin-strative restrictions for coxibs across the provinces.
We included the results of GPA use in this studyecause on introduction, the purported advantage of theoxibs was a lower incidence of severe gastrointestinalvents (bleeding and ulcers), as compared with non-elective non-steroidal anti-inflammatory drugs [29].
onsequently, it was assumed that GPA co-prescriptionould be reduced with coxibs, and this assumptionas been used in arguments for the cost-effectivenessf treatment with coxibs despite the incremental cost
ealth P
osgAipcituibrt
vpiMtfowvc
wMtaip(omtovotar
4
so
imEstic
abuttntbfcamTee
pttrtruat[oedBiatci
D.A. Marshall et al. / H
f coxibs compared to NSAIDs. Paradoxically, sub-equent studies have observed an increased rate ofastrointestinal events associated with coxib use [30].
previous analysis of the impact of the Ontario pol-cy restriction was restricted to a 6-month observationeriod post-coxib introduction and did not capture theoncurrent changes in GPA use [31]. Thus, we thoughtt important to track the use of GPAs during the sameime period to obtain the actual observed patterns ofse. If coxibs were being prescribed more frequentlyn place of NSAIDs, then a lower rate of GPA use woulde expected if clinicians believed that GPAs were notequired with coxibs. This would partly offset the addi-ional daily drug cost of coxibs compared to NSAIDs.
Our study spanned a period of 5 years. This pro-ides a sufficient number of observations in both there-coxib and post-coxib time periods to capture mean-ngful trends and avoid spurious spikes in the data.
amdani et al. [31] have previously reported a substan-ial and immediate impact of Ontario’s ‘limited use’ormulary listing for coxibs on the costs and quantityf overall NSAID prescriptions. No comparison groupas included in the analysis and the window of obser-ation post-coxib introduction was only 6 months afteroxibs were listed on ODB.
Our findings complement those of Fischer et al.ho undertook a similar analysis of the 50 stateedicaid agencies to examine the changes in prescrip-
ion patterns after the implementation of each prioruthorisation program [32,33]. For those states thatmplemented a prior authorisation program, the pro-ortion of coxib use decreased by an average of 11%measured as the proportion of defined daily dosesf coxibs to NSAIDs defined daily doses) from the 6onths before implementation to six months after, and
his varied depending on the degree of restrictivenessf the prior authorisation criteria. We have also shownariability in the effectiveness of policies dependingn the degree of restrictiveness. The prior authorisa-ion program in BC did indeed reduce coxib utilisation,nd this effect was greater than for the limited useestriction in ON.
.1. Study limitations
We used administrative billing records for pre-cribed drugs. These do not capture the use ofver-the-counter medications. Over-the-counter med-
4
i
olicy 84 (2007) 1–13 11
cations are a large component of the overall NSAIDarket, but limited to ibuprofen and aspirin in Canada.lderly patients who use NSAIDs regularly have atrong financial incentive to obtain them by prescrip-ion. As such, it is likely that a majority of NSAID usen this population is captured in the provincial druglaims databases.
A second limitation is that we used observationaldministrative databases, from which causation cannote directly determined, to retrospectively capture drugtilisation. There may be other factors that contributedo the pattern of coxib use, such as the underlying pat-erns of overall drug use in each province, that couldot be accounted for in this analysis with administra-ive data. Nonetheless, the strong temporal correlationetween the introduction of administrative restrictionsor coxibs, and the lack of other trends to explain thehange in NSAID utilisation, support our findings. Inll three jurisdictions, the policy environment for phar-aceuticals was stable over the observation period.here were no significant changes in the insurancenvironment that would be expected to have had anffect on uptake of NSAIDs.
The comparisons in this paper were limited to theatterns of NSAID utilisation amongst three provinceshat implemented a range of administrative policieso restrict the uptake of coxibs. Important questionsemain about the effect of the administrative restric-ions on patient outcomes—e.g., gastric, cardiac, andenal. This protective effect of coxibs has been shownsing administrative data in Ontario, but these studieslso found a significantly higher risk of upper gastroin-estinal bleeding for rofecoxib relative to celecoxib29]. In fact, somewhat paradoxically, the overall ratef hospitalisation for upper gastrointestinal bleeding inlderly residents of Ontario increased after the intro-uction of coxibs [30]. A similar effect was not seen inC [34]. Other observational studies have revealed an
ncreased risk of cardiac events (myocardial infarctionnd stroke) [7–12] and renal events [35–37]. We intendo evaluate the impact of administrative restrictions onlinical and economic outcomes associated with thentroduction of coxibs in a separate analysis.
.2. Conclusions
Under BC PharmaCare’s prior authorisation pol-cy, use of coxibs and total NSAIDS was substantially
1 ealth P
crutNBiaep
A
UNePiRybUtoc
R
[
[
[
[
[
[
[
[
[
[
[
[
[
[
2 D.A. Marshall et al. / H
urbed compared to Quebec. However, under the lessestrictive limited use policy in Ontario, the rate ofptake and use of coxibs was essentially the same ashat in Quebec, where there were no restrictions. TotalSAID use was greatest in Ontario compared to bothC and Quebec. In addition, the reduced use of cox-
bs observed in BC PharmaCare was partly offset byn increase in prescriptions through private insurancequal to about half of the increase observed in the otherrovinces.
cknowledgements
Contributors: Doreen Au, Faculty of Pharmacy,niversity of Toronto, assisted with data analysis;atalie Forde, Institute for Clinical Evaluative Sci-
nces, Toronto, Canada, assisted with data acquisition;aul Fortin, University of Toronto, assisted in obtain-
ng funding and conceptualising the research initially;ina Levya, Wyeth-Ayerst US, assisted with data anal-sis and provided research support (Note: Employedy the Centre for Evaluation of Medicines, McMasterniversity at the time of research); Steve Morgan, Cen-
re for Health Services and Policy Research, Universityf British Columbia, assisted in obtaining funding andonceptualising the research initially.
Source of support: CIHR Grant MOP-49454.
eferences
[1] Drug expenditure in Canada: 1985–2004. Ottawa: CanadianInstitute for Health Information; 2005.
[2] Drug expenditure in Canada: 1985–2003. Ottawa: CanadianInstitute for Health Information; 2004.
[3] Morgan S. Drug spending in Canada: recent trends and causes.Medical Care 2004;42:635–42.
[4] Bombardier C, Laine L, Reicin A, Shapiro D, Burgos-VargasR, Davis B, et al. Comparison of upper gastrointestinal toxicityof rofecoxib and naproxen in patients with rheumatoid arthritis.New England Journal of Medicine 2000;343:1520–8.
[5] Silverstein FE, Faich G, Goldstein JL, Simon LS, PincusT, Whelton A, et al. Gastrointestinal toxicity with celecoxibvs nonsteroidal anti-inflammatory drugs for osteoarthritis andrheumatoid arthritis: the CLASS study. A randomized con-
trolled trial. Celecoxib long-term Arthritis Safety Study. Journalof American Medical Association 2000;284:1247–55.[6] FitzGerald GA, Patrono C. The coxibs, selective inhibitorsof cyclooxygenase-2. New England Journal of Medicine2001;345:433–42.
[
olicy 84 (2007) 1–13
[7] Wright JM. The double-edged sword of COX-2 selectiveNSAIDs. Canadian Medical Association Journal 2002;167:1131–7.
[8] Mamdani M, Juurlink DN, Lee DS, Rochon PA, Kopp A, NaglieG, et al. Cyclo-oxygenase inhibitors versus non-selective non-steroidal anti-inflammatory drugs and congestive heart failureoutcomes in elderly patients: a population-based cohort study.Lancet 2004;363:1751–6.
[9] Mamdani M, Rochon P, Juurlink DN, Anderson GM, Kopp A,Naglie G, et al. Effect of selective cyclooxygenase-2 inhibitorsand naproxen on short-term risk of acute myocardial infarctionin the elderly. Archives of Internal Medicine 2003;163:481–6.
10] Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascularevents associated with selective COX-2 inhibitors. Journal ofAmerican Medical Association 2001;286:954–9.
11] Ray WA, Stein CM, Hall K, Daugherty JR, Griffin MR.Non-steroidal anti-inflammatory drugs and risk of seriouscoronary heart disease: an observational cohort study. Lancet2002;359:118–23.
12] Oberholzer-Gee F, Inamdar SN. Merck’s recall of rofecoxib—astrategic perspective. New England Journal of Medicine2004;351:2147–9.
13] Ontario Ministry of Health. Handbook of limited use drug prod-ucts. 38 ed. Ottawa: Ministry of Health; 2004.
14] BC PharmaCare, http://www.healthservices.gov.bc.ca/pharme/index.html. Special Authority Process Overview. Victoria:British Columbia Ministry of Health; c1995–2005 [updated2005 October 28; cited 2005 November 29]. Available from:http://www.healthservices.gov.bc.ca/pharme/sa/criteria/sachart.pdf.
15] WHO Collaborating Centre for Drug Statistics Methodology.ATC classification index with DDDs. 4 ed. Oslo, Norway:WHO; 2001.
16] Merlo J, Wessling A, Melander A. Comparison of dose standardunits for drug utilisation studies. European Journal of ClinicalPharmacology 1996;50:27–30.
17] Statistics Canada. Estimates of population by age group andsex, Canada, provinces and territories. CANSIM database[online]. Available from: http://cansim2.statcan.ca/ [accessed2003 November 11].
18] Brocklebank JC, Dickey DA. SAS System for Forecasting TimeSeries. Cary, NC: SAS Institute; 1986.
19] Ostrom CW. Time series analysis: regression techniques 2.Thousand Oaks, CA: Sage; 1990.
20] McDowall D, McCleary R, Meidinger EE, Hay RA. Interruptedtime series analysis. Thousand Oaks, CA: Sage; 1990.
21] Box GEP, Jenkins GM. Time series analysis: forecasting andcontrol. Oakland, CA: Holden-Day; 1976.
22] Newey WK, West KD. A simple, positive semi-definite,heteroskedasticity and autocorrelation consistent covariancematrix. Econometrica 1987;55:703–8.
23] Grimshaw JM, Thomas RE, MacLennan G, Fraser C, Ramsay
CR, Vale L, et al. Effectiveness and efficiency of guideline dis-semination and implementation strategies. Health TechnologyAssessment 2001;8:iii–iv.24] Dormuth CR, Maclure M, Bassett K, Jauca C, Whiteside C,Wright JM. Effect of periodic letters on evidence-based drug
ealth P
[
[
[
[
[
[
[
[
[
[
[
[
Journal of Medicine 2005;352:1081–91.
D.A. Marshall et al. / H
therapy on prescribing behaviour: a randomized trial. CanadianMedical Association Journal 2004;171(9):1057–61.
25] Therapeutics Initiative. Celecoxib (Celebrex): is it a break-through drug? Therapeutics Letter 1999;31.
26] Therapeutics Initiative. Selective COX-2 inhibitors: are theysafer NSAIDs? Therapeutics Letter 2002;39.
27] Therapeutics Initiative. COX-2 inhibitors update: do journalpublications tell the full story? Therapeutics Letter 2002;43.
28] Therapeutics Initiative. Rofecoxib (Vioxx®) withdrawal gen-erates uncertainty about “COX-2s”: do product monographsadequately inform? Therapeutics Letter 2004;53.
29] Mamdani M, Rochon PA, Juurlink DN, Kopp A, Anderson GM,Naglie G, et al. Observational study of upper gastrointestinalhaemorrhage in elderly patients given selective cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs. British Medical Journal 2002;325:624–7.
30] Mamdani M, Juurlink DN, Kopp A, Naglie G, Austin PC,Laupacis A. Gastrointestinal bleeding after introduction ofCOX 2 inhibitors: ecological study. British Medical Journal
2004;328:1415–6.31] Mamdani M, Rochon P, Laupacis A, Anderson G. Initial pat-terns of use of COX-2 inhibitors by elderly patients in Ontario:findings and implications. Canadian Medical Association Jour-nal 2002;167:1125–6.
[
olicy 84 (2007) 1–13 13
32] Fischer MA, Schneeweiss S, Avorn J, Solomon DH. Medi-caid prior-authorization programs and the use of cyclooxy-genase-2 inhibitors. New England Journal of Medicine2004;351:2187–94.
33] Hamel MB, Epstein AM. Prior-authorization programs for con-trolling drug spending [comment]. New England Journal ofMedicine 2004;351(November (18)):2156–8.
34] Mamdani M, Warren L, Kopp A, Paterson JM, Laupacis A,Bassett K, et al. Changes in rates of upper gastrointestinal hem-orrhage after the introduction of cyclooxygenase-2 inhibitors inBritish Columbia and Ontario. Canadian Medical AssociationJournal 2006;175:1535–8.
35] Cheng HF, Harris RC. Renal effects of non-steroidal anti-inflammatory drugs and selective cyclooxygenase-2 inhibitors.Current Pharmaceutical Design 2005;11:1795–804.
36] Nussmeier NA, Whelton AA, Brown MT, Langford RM, HoeftA, Parlow JL, et al. Complications of the COX-2 inhibitorsparecoxib and valdecoxib after cardiac surgery. New England
37] Barkin RL, Buvanendran A. Focus on the COX-1 and COX-2 agents: renal events of nonsteroidal and anti-inflammatorydrugs-NSAIDs. American Journal of Therapeutics 2004;11:124–9.
