Cardiovascular events associated with the use of four nonselective NSAIDs (Etodolac, Nabumetone, Ibuprofen, or Naproxen) Versus a Cyclooxygenase-2 Inhibitor (Celecoxib): A population-based analysis in taiwanese adults

:

i

:

ClinicalTherapeutics/Volume

28, Number 11' 2006

. . . .

Cardiovascular Events Associated with the Use of Four Nonselective NSAIDs (Etodolac, Nabumetone, Ibuprofen, or Naproxen) Versus a Cyclooxygenase-2 Inhibitor (Celecoxib): A Population-Based Analysis in Taiwanese Adults W e n g - F o u n g H u a n g , PhD1; Fei-Yuan Hsiao, MS2; Y u - W e n W e n PhD3; a n d Yi-Wen Tsai, P h D 3

l lnstitute of Health and Welfare Policy, National Yang-Ming University, Taipei, Taiwan; 2Division of Health and Welfare Policy Management, Institute of Public Health, National Yang-Ming University, Taipei, Taiwan; and 3Division of Health Policy Research, National Health Research Institutes, Miaoli, Taiwan ABSTRACT Background: Serious cardiovascular events (CVEs) have been linked to the use of cyclooxygenase (COX)-2 inhibitors, a category of selective NSAIDs. However, few studies are available that have compared the risk for CVEs between COX-2 inhibitors and nonselective NSAIDs in adults undergoing long-term treatment. Objectives: The present study assessed (1) whether long-term use of nonselective NSAIDs (etodolac, nabumetone, ibuprofen, or naproxen) is associated with an increased risk for treatment-related CVEs (acute myocardial infarction [AMI], angina, cerebrovascular attack [CVA], and/or transient ischemic attack [TIA]) compared with long-term use of celecoxib and (2) which factors are associated with the risk for treatment-related CVEs in long-term users of nonselective NSAIDs in Taiwan. Methods: This population-based analysis used data from the Taiwanese Bureau of National Health Insurance (Taipei, Taiwan) database. Eligible patients were aged _>18 years and had been receiving etodolac, nabumetone, ibuprofen, naproxen, or celecoxib for _>180 days between January 1, 2001, and December 31, 2003. The primary outcomes measure was the prevalence of serious CVEs (AMI, angina, CVA, and/or TIA requiring hospitalization) after initiation of treatment. Analyses were performed on data from all eligible patients; person-time exposures to the drugs and hazard ratios (HRs) were calculated to determine the risk for CVEs with long-term use. Results: A total of 16,326 patients (8166 men, 8160 women; mean [SD] age, 61.83 [20.23] years) who had received long-term treatment with etodolac (2014 [12.34%]), nabumetone (2262 [13.86%]), ibuprofen (5239 [32.09%]), naproxen (3049

[18.68%]), or celecoxib (3762 [23.04%]) were identified. The overall prevalences of AMI, angina, CVA, and TIA were higher in long-term users with a history of cardiovascular disease (CVD) than in those without (AMI, 4.76% vs 0.99%; angina, 4.11% vs 0.43%; CVA, 7.74% vs 1.51%; and TIA, 4.03% vs 0.52%) (all, P < 0.01). The HRs for AMI, angina, CVA, and TIA were not significantly different between the NSAID and celecoxib groups. History of CVD played a significant role in recurrence during the period studied; the HRs (95% CIs) were 2.29 (1.22-4.32) for AMI, 6.19 (3.56-10.78) for angina, 3.56 (2.80-4.52) for CVA, and 6.60 (3.72-11.73) for TIA. Preexisting medical conditions (hypertension, dyslipidemia, diabetes mellitus, congestive heart failure, chronic renal disease) also significantly affected the risk for CVEs. Conclusions: In this cohort study of long-term (>180 days) use of NSAIDs in Taiwanese adults, no significant differences in the risk for treatment-related CVEs were observed between groups prescribed 1 of 4 nonselective NSAIDs (etodolac, nabumetone, ibuprofen, or naproxen) or celecoxib. History of CVD was the most significant determinant of CVE risk. Patients with preexisting medical conditions appeared to have a significantly higher risk for CVEs associated with the use of NSAIDs and celecoxib compared with patients without these conditions. (Clin Ther. 2006;28:1827-1836) Copyright © 2006 Excerpta Medica, Inc.

AcceptedflorpublicationSeptember14, 2006. doi:l 0.1016/j.clinthera.2006.11.009 0149-2918/06/$19.00 Printed in the USA.Reproductionin wholeor part is not permitted. Copyright© 2006 ExcerptaMedica,inc.

ClinicalThera~ peutlcs , i : : : .

.

.

.

.

.

.

.

.

.

Key words: adverse drug events, celecoxib, NSAIDs, data analysis.

INTRODUCTION Cyclooxygenase (COX)-2 inhibitors form a category of selective NSAIDs that have been associated with fewer gastrointestinal adverse events (AEs) compared with nonselective NSAIDs. 1,2 Treatment-related cardiovascular events (CVEs) had been associated with the use of COX-2 inhibitors even before the withdrawal of rofecoxib from the market on September 30, 2004. `3-8 That withdrawal stirred further discussion regarding a possible class effect of COX-2 inhibitors, with such concerns extending to nonselective NSAIDs. 9-15 Two federal US government-sponsored studies involving COX-2 inhibitors and nonselective NSAIDs were halted due to concern over treatmentrelated CVEs. 16 On April 6, 2005, the US Food and Drug Administration's (FDA's) Decision Memo on NSAIDs suggested that the available data did not permit a rank ordering of COX-2 selective NSAIDs with regard to CVEsJT: data from large, long-term, controlled clinical trials that included a comparison of COX-2 selective with nonselective NSAIDs did not clearly find that the COX-2 selective agents conferred a greater risk for serious treatment-related CVEs than nonselective NSAIDs. The FDA's memo further stated that the available data were best interpreted as being consistent with a class effect of an increased risk for serious treatment-related CVEs for COX-2 selective and nonselective NSAIDs. 17 Findings from more recent studies, 18-26 although inconsistent, have provided more evidence to support these statements. For example, the results of a study22 of Medicaid data that comprised 1005 patients using COX-2 inhibitors and 5245 patients using a non-naproxen NSAID did not find a difference in the rate of CVEs between the 2 groups. Nonetheless, the existence of a class effect in celecoxib and/or other NSAIDs was still questioned. To our knowledge, no studies in long-term users of 4 NSAIDs and celecoxib using one survival model are available. Only 2 COX-2 inhibitors (rofecoxib and celecoxib) were commercially available in Taiwan at the time rofecoxib was withdrawn from the market. On July 1, 2001, the Taiwanese Bureau of National Health Insurance (BNHI) (Taipei, Taiwan) began covering the cost of celecoxib. In a previous retrospective cohort study27 that included Taiwanese patients who received

1828

,

.

.

.

::

.

:;

.

.

.

.

.

:

.

.

.

:

.

:

long-term treatment of celecoxib (n = 3762 [39.2%]), rofecoxib (n = 1550 [16.1%]), or meloxicam (n = 4290 [44.7%]), we found that patients who received celecoxib had a significantly lower risk for treatmentrelated CVEs than those receiving meloxicam. Celecoxib users had a significantly lower hazard ratio (HR) for acute myocardial infarction (AMI) (HR, 0.78; 95% CI, 0.63-0.96; P = 0.02) and cerebrovascular accident (CVA) (HR, 0.81; 95% CI, 0.70-0.93; P < 0.001). Using population-based data from the BNHI for January 1, 2001, to December 31, 2003, the present study assessed (1) whether long-term use of nonselective NSAIDs (etodolac, nabumetone, ibuprofen, or naproxen) is associated with an increased risk for treatment-related CVEs (AMI, angina, CVA, and/or transient ischemic attack [TIA]) compared with long-term use of celecoxib and (2) which factors are associated with the risk for treatment-related CVEs in long-term users of nonselective NSAIDs in Taiwan.

MATERIALS AND METHODS Study Population For this objective study, data were analyzed from Taiwanese adult (aged >18 years) patients, for whom the medical data were obtained from the BNHI, which covered medical expenses for _>99% of Taiwan's population of 23 million from 2001 through 2003. The National Health Insurance (NHI) databases used a population-based data-collection process. A variety of manually maintained outpatient and inpatient medical records were also included in the NHI databases. Information on diagnosis, treatment, and the occurrence of clinical AEs has been tracked with identification numbers since 1996. Therefore, the NHI data files allowed for cohort identification, classification of CVE risk factor status, and end point determination. Eligible patients included all enrollees in the BNHI with pharmacy records indicating continuous use (_>180 days) of I of the 5 study drugs between January 1, 2001, and December 31, 2003. It was assumed that patients who had records of having a prescription filled were using the drug. We did not differentiate the dosage of celecoxib because preliminary analysis of our study population found that 92% of celecoxib users were prescribed a daily dose of 200 mg (78.9%) or less (13.5%). The exclusion criteria were as follows: any concomitant use of 1 of the 5 medications studied or

i

'

~

~,

Volume 28 Number 11

i other NSAIDs; any lag of >14 days between 2 prescription fill dates, which suggested discontinuity of use; and duration of use of a study drug of <180 days, as this would not be considered long-term use. Patient identification numbers on all claims data released by BNHI were scrambled to safeguard the confidentiality of individually identifiable information. This study, therefore, did not require informed consent or institutional review board approval.

Data Collection For each of the 5 study drugs prescribed during an outpatient visit, the claims data included starting date (date prescription was dispensed), quantity dispensed, dose prescribed, and prescription duration. We defined the end date of prescription duration as the prescription dispensing date plus prescription duration. We created a prescription profile for all included patients for further examination purposes. To identify any preexisting (within 1 year before the initiation of treatment) clinical conditions and the first-time occurrence of the CVEs analyzed in the present study, we also created a medical history profile for each patient using 2 subsets of the NHI databases that provided data from outpatient and inpatient visits. The primary outcome of interest was the prevalence of serious CVEs (AMI, angina, CVA, and/or TIA requiring hospitalization) that occurred after the initiation of treatment. We defined each case, based on the

International Classificationof Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes,28 as AMI (410.xx and 411.xx), angina (413.xx and 414.xx), CVA (433.xx and 444.xx), or TIA (435.xx and 437.1). The covariate variables included age, sex, accumulated duration of use based on prescription data, and preexisting (in the year before treatment initiation) CVEs and/or hypertension (ICD-9-CM codes 401.xx-405.xx), dyslipidemia (272.4), diabetes mellitus (250.xx), congestive heart failure (CHF) (428.xx), and/or chronic renal disease (580.xx-587.xx) identified using outpatient or inpatient records. Patients were defined as having cardiovascular disease (CVD) if any inpatient visit record indicated the presence of a CVE <1 year before treatment with a study drug was initiated.

Statistical Analysis We used descriptive statistics to compare baseline characteristics between treatment groups. The ~2 test

:November2006

~i:

i: ::i •

i

:

W;-F. Huang et at.

and 1-way analysis of variance test were used to perform the comparisons between multiple stratified groups. Statistical significance for all analyses was defined as P < 0.05. A Cox proportional hazards model was used for assessing the association of long-term NSAID treatment with the occurrence or recurrence of AMI, angina, CVA, and/or TIA. The date of the first outpatient visit at which any 1 of the 5 study medications was dispensed was considered as the day of treatment initiation. Duration of follow-up was considered as the date of treatment initiation until the first occurrence or recurrence of a CVE, or, in those who did not experience a CVE, the end of the period studied. The Cox proportional hazards model was performed with S-PLUS version 7.0.3 (Insightful Corporation, Seattle, Washington); all other analyses were conducted using SAS version 8.2 (SAS Institute Inc., Cary, North Carolina).

RES U LTS Data from 16,326 patients (8166 men, 8160 women; mean [SD] age, 61.83 [20.23] years) who received >180 days of treatment with etodolac (2014 [12.34%]), nabumetone (2262 [13.86%]), ibuprofen (5239 [32.09%]), naproxen (3049 [18.68%]), or celecoxib (3762 [23.04%]) were included (Table I). Total duration of follow-up was 139,560 person-years. A total of 56.01% (9145/16,326) of the patients were aged >65 years. The mean (SD) duration of continuous use during the period studied was 385.51 (211.05) days. The mean (SD) durations of nabumetone use (424.58 [234.15]) and naproxen use (417.54 [240.15]) were significantly longer than those of the other 3 medications (etodolac, 395.61 [207.48]; ibuprofen, 347.54 [200.22]; and celecoxib, 383.52 [175.93]) (both, P < 0.01 vs etodolac, ibuprofen, and celecoxib). The nabumetone and naproxen cohorts each contained a significantly large proportion of patients whose duration of continuous use was >540 days (26.26% [594/2262] and 25.22% [769/3049], respectively; P < 0.001). Among the study patients, there were 210 (1.29%) with a history of AMI; 389 (2.38%), angina; 1137 (6.96%), CVA; and 248 (1.52%), TIA. The overall prevalences of AMI, angina, CVA, and TIA were higher in long-term users with a history of CVD than in those without (AMI, 4.76% vs 0.99%; angina, 4.11% vs 0.43%; CVA, 7.74% vs 1.51%; and TIA, 4.03% vs 0.52%; all, P < 0.01). As for preexisting conditions,

,,~

T829

C

=

~

.

.

.

~

~

o~

~

~

--II

~

~

~

~

~

'

o

°6.

~

~

..

.

.

.

.~

~

~

. ~

!~ , -

~

E

E

EEES

o o o o 6 6 6 6

o ~ S E o o o o o o 6 6 6 6 d

v

v

v

v v v ~

v

o o

o o

~

~

~

o o

~

~

~

~

~

~

~

v

v

~

v

! ~ I 0 I ~ v

l

~

~

~._

m

>

0

~

~11

~.~

"

E

Z ~'~

"

o

* ~

=

.

~

~

.--

~

0

~

~ 0

~

~

~

~

~

~

~ 0 0 ~

~

~

0

II

~

~

~

u ~

~

~E

0

~

~ 0 ~

~

~

~

~

~ 0

~

~

~

~ ~ ~

~

~ ~~

~

~

~

~

~ ~~

~

~

~

o

~

.-

:: •

;, i

IL:: ~

:::::

::

~ " ~- C ~ 0 0 E'~ ~ 7

0

~

~

~

.... •

~

II ~ ~ < ~ ~ ~\o ~ 0 ~

~c~

:tS3o

c

~

~

O~

~

o

!Voiu~e28Ni.,,,.~ri~i

~

~

0

=

•

: W.,F. ."uang et al:

:

5994 (36.71%) patients had current hypertension; 843 (5.16%), dyslipidemia; 2544 (15.58%), diabetes mellitus; 543 (3.33%), CHF; and 646 (3.96%), chronic renal disease. The etodolac, nabumetone, and celecoxib groups had significantly higher rates of preexisting hypertension (etodolac, 47.22 % [951/2014]; nabumetone, 46.64% [1055/2262]; and celecoxib, 47.16% [1776/3762]) compared with users of ibuprofen (23.00% [1205/5239]) and naproxen (33.09% [1009/3049]) (all, P < 0.01 vs ibuprofen and naproxen). Similarly, the etodolac, nabumetone, and celecoxib groups had significantly higher rates of diabetes mellitus (etodolac, 20.41% [411/2014]; nabumetone, 21.40% [484/2262]; and celecoxib, 19.94% [749/3762]) compared with the groups using ibuprofen (8.82% [462/5239]) and naproxen (14.33% [437/3049]) (all, P < 0.01 vs ibuprofen and naproxen). As for the presence of dyslipidemia, the nabumetone and celecoxib groups had significantly higher rates (nabumetone, 6.50% [147/2262]; celecoxib, 7.02% [263/3762]) compared with the groups that used ibuprofen (3.02% [157/5239]) and naproxen (4.92% [149/3049]) (both, P < 0.01 vs ibuprofen and naproxen). During the period studied, 169 (1.03%) patients experienced AMI; 84 (0.51%), angina; 317 (1.94%), CVA; and 93 (0.57%), TIA. Table II stratifies the occurrence of CVEs by medication based on the patients' history of each event. The data suggested that during the period studied, the prevalence of AMI in those with a history of AMI was 4.76% (10/210), whereas in 0.99% (159/16,116) of those without a history, AMI developed (P < 0.01). In those without a history of AMI, the rate of occurrence of AMI during the period studied was highest among users of nabumetone (1.43 % [32/2231]), followed by users of etodolac (1.11% [22/1976]), naproxen (1.03% [31/3018]), celecoxib (0.92% [34/3685]), and ibuprofen (0.77% [40/5206]), but there were no significant differences between groups. In patients with a history of AMI, the rate of recurrence was highest in users of nabumetone (9.68% [3/31]), followed by users of naproxen (6.45% [2/31]), celecoxib (3.90% [3/77]), ibuprofen (3.03% [1/33]), and etodolac (2.63% [1/38]). Patients with a history of angina who were using ibuprofen had the highest recurrence rate (5.45% [3/55]), followed by users of celecoxib (4.69% [6/128]), etodolac (3.85% [3/78]), naproxen (3.33% [2/60]), and nabumetone

(2.94% [2/68]). Patients with a history of CVA who were using naproxen had the highest rate of recurrence (11.76% [16/136]), followed by users of etodolac (8.65% [23/266]), nabumetone (6.93% [14/202]), ibuprofen (6.78% [12/177]), and celecoxib (6.46% [23/256]). Patients with a history of TIA who were using nabumetone had the highest rate of recurrence (5.41% [2/37]), followed by users of naproxen (5.00% [2/40]), celecoxib (4.55% [4/88]), etodolac (4.17% [2/48]), and ibuprofen (0) (Table II). Table III shows the results of the survival analysis for each CVE, based on age, sex, medication prescribed, prescription duration, history of CVD, and preexisting medical conditions. The risk for recurrence of AMI was similar between the groups of nonselective NSAID users and celecoxib. History of AMI was significantly associated with recurrence of AMI during the period studied (HR, 2.29; 95% CI, 1.22-4.32; P -- 0.01). Patients with hypertension (HR, 1.41; 95% CI, 1.04-1.92; P -- 0.03), diabetes mellitus (HR, 1.62; 95% CI, 1.18-2.23; P < 0.01), CHF (HR, 2.17; 95% CI, 1.38-3.39; P < 0.01), or chronic renal disease (HR, 1.81; 95% CI, 1.17-2.81; P -- 0.01) were at a significantly greater risk for AMI compared with those without a history. History of angina was significantly associated with recurrence during the period studied (HR, 6.19; 95% CI, 3.56-10.78; P < 0.01). Patients with dyslipidemia (HR, 2.79; 95% CI, 1.60-4.86; P < 0.01) or CHF (HR, 2.04; 95% CI, 1.08-3.85; P -- 0.03) were at a significantly greater risk for angina. History of CVA was also significantly associated with recurrence during the period studied (HR, 3.56; 95% CI, 2.80-4.52; P < 0.01). Patients with diabetes mellitus (HR, 1.41; 95% CI, 1.11-1.78; P < 0.01) or CHF (HR, 1.46; 95% CI, 1.01-2.10; P = 0.04) had a significantly higher risk for CVA during the period studied. Similarly, a history of TIA was significantly associated with recurrence (HR, 6.60; 95% CI, 3.72-11.73; P < 0.01). Also, patients with diabetes mellitus (HR, 1.74; 95% CI, 1.15-2.64; P = 0.01) had higher risk for TIA during the period studied. DISCUSSION

The treatment-related AEs associated with the use of COX-2 inhibitors and nonselective NSAIDs have evolved into an important drug-tolerability issue. The FDA's Decision Memo on April 6, 2005, provided an updated benchmarking of NSAID tolerability, both for COX-2 inhibitors and nonselective NSAIDs. 17 In

r~ I

~

~'-- ',.0

'~- C~

CO,--

~.O C'.I

cr~ . ' ~

O0

0 ~

0

r

,q

O0 ~

E

:

_~ "'~

:~

u

.~- ~_ I

0

~

~

.

"~-o ~"

~

or)

0

~ x

~

~,~

~

co h. ~

~

"~o

oo ~.. ~

,_

~:.

oo

0

~

oo

~

~l'..

~

~

~

~

u"~

0

~,~

~

O

.

.r-1

p~ ~

~.,-.

I'~. 0

~

oo ~-.

.

"--o

~-~

oo h.

~

,:--

oo

~

G'~ ~

--~

i~.

0

eq

u~

P~ 0

r-

~

~

~

II

u

t.~ Z

x

c~

0

[5a~

Z

~0

0 '~-

o

b ~'~

o IZ

~

'.~

0",

~

m

c~ ~

, l~

,--

.

oO c'.l

~-~

~

.

r< d .

{q

~

~

.

~

~

.

r~ o~ .

".D "~-

~

.._.,

~

~

.

~

~

~

~

0 0 . q 0.

~

C~I ,r'-

~

~'~

~

m

v..o

~

~.~.

b~~

"-'~

o

h:~

~"-"

-~

b~

_o o ~-.~

Z

, ,~ ~.

'--

v

~

0o~"

oo

oo--

,1

o~"

~

o -~

> U

~

~

~o ,,D

~o~

~

~ o~" ~

~

u

~

~~-

,~"

~

~"

oo~"

o

-~

~

~-~

~

~

~ "

~

~ "

,~-

~c~ ~d

~

~-~.

~-~-"

~

oo

~ o~

~0

oO ,o

"U ~

,m

r-.~

,~

~"

~-~

u'}

,. ~1 ,-

,g~

c~

o0

~

o

r~ ~

o {N

~

0"~

~

0

w

0

m o~

,--

I'-,

~

Oh

0

~

~ .

~

E

~ o

oo Oh

'

o ~

8 b

i

ooo

>~

~.

~-o

~ "~-o

r-.. ~ m -~ ~

~

~

kC)

0 -r-

~ ~'-

~

~

~

~

~

~

~"

i

~:~

~

oo

~

~

~-

o

~-.~

~-

i

~

.

0

I'-,

oo

Oh

0

0

i

~: -,s

i

e~ b

'.

>~

b~_ o o~

~

E U

~

~-.

_

'u .

z ¢

.-

-~

~

oo ~

~

I'.,,

o ~

~.

M3

',C~ I ' - ,

.b~--.~

.

o •

;

~ .~ ,,>, ui<

e

,-~"' <

e

X , , >, u

-

e

_<">' ~-

e

e-

2

,-0 0 V

O0 .~. 0

~0

oo

I~ 0

0

0

0

0

0'~ 0

~-" CO

dc~ddd

0 V

u

..O e"

o

0 ~

<

0 0

VVV

0'~

O

~ 0

0

Ox

~ ~

0 ~

~

,'-"

0

d

d

d

'q" 0

'q" ~'--

~

V

0

~

0

m

E

E

V

O 0

~-- ,~- t.D ~,O ,"- ~ 0 kO

('N I

~'N I

("N I

('N I

~m --

~'--

~

o

~

d d ~ d d O

v

r'~ 35

d

d

g

,-': d

~

0

< Z

~ 0

Ck V

C)

~ 0

0 0

0

0

d d d d d

0 V

~--

~

~

~

~

0

0

0

u

V

0

~

c

¢._u c

I I

O c-

0

~

0

.

0

.

.

.

.

e-

~

II ,.,/

O

o

u

C~ !

o._~

N

E

~

~ 0

O

0

~

O ~ 0 0 0

r~

O 0

0 0 V

V Ii

u ~

0

~

,e-- v

g

m oO

t'N

t"-I

0

I

I

1

~

~

1

I

g

B.~.

~.~:

-g

0

I

I

d

I

d

~

I

d

g

I

II <

0 c ~'~

II

~ c

e-

g d

l

V

E ~5

0

0

0

~

0

0

0

0

0 0 0 0 0 V V

0

O ,

~o .2 -~>.

I I I I ~

I

&4%4g

<

~

0

~

o d d d ~ u'3

C',I

~

~

X

0

0 ~ 0 0 0

T

55 ~

~

.--

C 0

¢..

~

O "~

. -~"

~ _ ~ . ~O

0

E

..0

0

-%: C 0

O_

,.2"

-

m

"~

o

o

o E

_O ..O

"r-

~Z N

~ o.

0

E

o

~ ,,

~~

~ge

meloxicam. Similar results were found in a case-control study22 (1718 case-patients with new, nonfatal AMI admitted to 1 of 36 hospitals in a 5-county area of the United States and 6800 controls randomly selected from the same counties) that evaluated the effects of COX-2 inhibitors on the risk for nonfatal AMI using 1 conventional NSAID as the comparator. While the class effects of both coxibs and nonselective NSAIDs for CVEs have become known, the risks vary depending on the history of the patient and his or her concurrent medications) 8-~° Nonetheless, a difference in CVE class effects between COX-2 inhibitors and NSAIDs is still being questioned. We chose celecoxib as the comparator to address this issue.

the present study, 3 years of Taiwanese national insurance claims data (January 1, 2001-December 31, 2003) were used to construct a national cohort and provide long-term information from prescription-drug users. We found no significant difference in rates of treatmentrelated CVEs between the users of the 4 NSAIDs and celecoxib. History of CVD, the recurrence of a preexisting CVE, and a preexisting (within 1 year before initiation of treatment) medical condition were the most significant determinants of such risks. Although package inserts for COX-2 inhibitors include a warning for patients with a history of CVD, physicians may not be fully aware of the risks associated with prescribing COX-2 inhibitors or nonselective NSAIDs. In addition, COX-2 inhibitors have stirred comprehensive debate concerning the advantage of this subclass in reducing the risk for AEs (ie, gastrointestinal AEs) versus the occurrence of serious AEs (ie, CVEs).29'30 Using prescription event-monitoring methodology from the New Zealand Intensive Medicines Monitoring Program, a prospective, longitudinal, observational cohort study to identify CVEs associated with the use of celecoxib versus rofecoxib was conducted. 2s In that study, 4882 (18%) and 6267 (19%) patients who received >1 prescription for rofecoxib or celecoxib, respectively, between December 1, 2000, and November 30, 2001, and completed a follow-up questionnaire by September 2004 were included. After an adjustment for age, the HR for thrombotic CVEs was 0.94 (95% CI, 0.51-1.70) for celecoxib compared with rofecoxib. As a result, in an interim analysis of the Intensive Medicines Monitoring Program, data suggested that in postmarketing use in routine clinical practice in New Zealand, there was no significant difference in the risk for thrombotic CVEs in patients receiving celecoxib compared with those receiving rofecoxib. However, in our previous report, 27 a retrospective, observational study which included 3 cohorts of Taiwanese patients who received celecoxib (3762 [39.2%]), rofecoxib (1550 [16.1%]), or meloxicam (4290 [44.7%]), found that patients receiving celecoxib had a significantly lower risk for CVEs than those receiving meloxicam. Celecoxib users had significantly lower HRs for developing AMI (HR, 0.78; 95% CI, 0.63-0.96; P = 0.02) and CVA (HR, 0.81; 95% CI, 0.70-0.93; P < 0.001) compared with meloxicam users. However, patients who received rofecoxib were not found to be at a higher risk for treatment-related CVEs than those who received

1834

~:• ~ ~ i:~:

~:~

.....

~

::,i::

Study Limitations The clinical diagnoses in this data analysis were not validated using medical record reviews. Patterns of discontinued use of prescription medication due to CVEs was not studied. Although we adjusted the methods to cover a range of potential CVE risk factors, there were several variables in the BNHI database that we did not record for study purposes, including family history of CVD, weight, and smoking habit. In addition, it may be difficult to control the potential confounding factor of patients using nonprescribed treatment with aspirin or another NSAID during the period studied. The databases contained records of only those medications covered by the BNHI. As a result, data concerning drugs not covered by the BNHI, such as over-the-counter drugs, were not included. Information on treatment compliance was not available for assessment. Because of the limited period studied, only CVEs that occurred within 1 year before treatment initiation were recorded; therefore, some selection bias might have existed. As we did not use data from the National Mortality File database, we did not know whether any fatal CVEs were associated with long-term NSAID treatment. There was no NSAID-naive control group in the study design, which might have clarified any differences in CVE class effects between COX-2 inhibitors and NSAIDs. Although the present study provided observational data concerning the association of the use of nonselective NSAIDs and the COX-2 inhibitor celecoxib with CVEs, we believe that it is important to continue to explore potential risks associated with the use of nonselective NSAIDs and COX-2 inhibitors in prospective, randomized, controlled, long-term studies.

~

: :

,:~'~ :~vo~ume

W.-F. Huang et aL CONCLUSIONS In this cohort study of long-term (_>180 days) use of NSAIDs in Taiwanese adults, no significant differences in treatment-related CVEs were observed between groups prescribed 1 of 4 nonselective NSAIDs (etodolac, nabumetone, ibuprofen, or naproxen) or celecoxib. History of CVD was the most significant determinant of CVE risk. Patients with preexisting medical conditions (eg, diabetes mellitus, CHF, dyslipidemia) appeared to have a significantly higher risk for CVEs associated with the use of NSAIDs and celecoxib compared with patients without these conditions. ACKNOWLEDGMENTS The authors thank the BNHI, Department of Health, and National Health Research Institutes (all, Taipei, Taiwan) for providing access to the BNHI databases. The authors also thank Chao-Ming Huang, MS (BNHI), for his support in retrieving data and for his suggestions regarding the structure of data from the NHI. Dr. Huang has received a research grant from Merck Foundation for an unrelated study. He has also represented Taiwanese government investment in serving on the boards of directors at 2 biotechnology companies in Taiwan.

9.

10. 11. 12. 13. 14. 15.

16. 17.

18.

REFERENCES 1. FitzGerald GA, Patrono C. The coxibs, selective inhibitors ofcyclooxygenase-2. N EnglJ Med. 2001 ;345:433-442. 2. Gajraj NM. Cyclooxygenase-2 inhibitors. AnestkAnalg. 2003; 96:1720-1738. 3. Adverse Drug Reactions Advisory Committee. RoFecoxib, celecoxib, and cardiovascular risks. AustAdv Drug Reactions Bull. 2003;22:19. 4. Graham DJ, Campen D, Cheetham C, et al. Risk of acute cardiac events among patients treated with cyclooxygenase-2 selective and non-selective nonsteroidal anti-inflammatory drugs [abstract]. PharmacoepidemiolDrugSaf 2004;13:$287$288. 5. Ray WA, Stein CM, DaughertyJR, et al. COX-2 selective non-steroidal anti-inflammatory drugs and risk of serious coronary heart disease. Lancet. 2002;360:1071-1073. 6. Day R. Hypertension in the patient with arthritis: Have we been underestimating its significance?J Rheumatol. 2003; 30:642-645. 7. Ray WA, Griffin MR, Stein CM. Cardiovascular toxicity of valdecoxib. N EnglJ Med. 2004;351:2767. 8. Farkouh ME, Kirshner H, Harrington RA, et al, For the TARGET Study Group. Comparison of lumiracoxib with naproxen and ibuprofen in the Therapeutic Arthritis

19.

20.

21.

22.

23.

Research and Gastrointestinal Event Trial (TARGET), cardiovascular outcomes: Randomised controlled trial. Lancet. 2004;364:675-684. Finckh A, Aronson MD. Cardiovascular risks of cyclooxygenase-2 inhibitors: Where we stand now. Ann Intern Med. 2005;142:212-214. Edwards IR. What are the real lessons From Vioxx? Drug Saf 2005;28:651 658. Topoi EJ. Arthritis medicines and cardiovascular events"house of coxibs."JAMA. 2005;293:366-368. Topoi EJ, Falk GW. A coxib a day won't keep the doctor away. Lancet. 2004;364:639-640. Hampton 3-. Experts point to lessons learned From controversy over rofecoxib safety.JAMA. 2005;293:413-414. Bannwarth B. Do selective cyclo-oxygenease-2 inhibitors have a Future? DrugSaf 2005;28:183-189. Kuehn B. FDA panel: Keep COX-2 drugs on market: Black box for COX-2 labels, caution urged For all NSAIDs. JAMA. 2005;293:1571-1572. Hampton T. Officials halt NSAID prevention trials.JAMA. 2005;293:664-665. US Food and Drug Administration. Analysis and recommendations For agency action-COX-2 selective and nonselective NSAIDs. Available at: http://www.fda.gov/ cder/drug/infopage/COX2/default.htm. Accessed August 10, 2005. Solomon SD, McMurray JJ, Pfeffer MA, et al, For the Adenoma Prevention with Celecoxib (APC) Study Investigators. Cardiovascular risk associated with celecoxib in a clinical trial For colorectal adenoma prevention. N EnglJ Med. 2005;352:1071-1080. Bresalier RS, Sandier RS, Quan H, et al, For the Adenomatous Polyp Prevention on Vioxx (APPROVe) Trial Investigators. Cardiovascular events associated with rofecoxib in a colorectat adenoma chemoprevention trial [published correction appears in N EnglJ Med. 2006;355: 221 ]. N EnglJ Med. 2005;352:1092-1102. Nussmeier NA, Whelton AA, Brown MT, et al. Complications of the COX-2 inhibitors parecoxib and valdecoxib after cardiac surgery. N EnglJ Med. 2005;352: 1081-1091. Shaya FT, Blume SW, Blanchette CM, et al. Selective cyclooxygenase-2 inhibition and cardiovascular effects: An observational study of a Medicaid population. Arch Intern Med. 2005;165:181-186. Kimmel SE, Berlin JA, Reilly M. Patients exposed to rofecoxib and celecoxib have different odds of nonfatal myocardial infarction. Ann Intern Med. 2005;142:157-164. Graham DJ, Campen D, Hui R, et al. Risk of acute myocardial infarction and sudden cardiac death in patients treated with cyclo-oxygenase 2 selective and non-selective non-steroidal anti-inflammatory drugs: Nested case-control study. Lancet. 2005;365:475-481.

24. Bernatsky S, Hudson M, Suissa S. Anti-rheumatic drug use and risk of hospitalization for congestive heart failure in rheumatoid arthritis. Rheumatology(Oxford). 2005;44:677680. 25. Harrison-Woolrych M, Herbison P, McLean R, et al. Incidence of thrombotic cardiovascular events in patients taking celecoxib compared with those taking rofecoxib: Interim results from the New Zealand Intensive Medicines Monitoring Programme. DrugSaf 2005;28:435442. 26. Kasliwal R, Layton D, Harris S, et al. A comparison of reported gastrointestinal and thromboembolic events between roFecoxib and celecoxib using observational data. Drug Saf. 2005;28:803-816. 27. HuangWF, Hsiao FY,Tsai YW, et al. Cardiovascular events associated with long-term use of celecoxib, rofecoxib and meloxicam in Taiwan: An observational study. Drug Sag 2006;29:261-272. 28. US Dept of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics (NCHS).

International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Hyattsville, Md: NCHS; 2003. 29. Topoi EJ. Failing the public healthrofecoxib, Merck, and the FDA. NEngl J Med. 2004;351:1707-1709. 30. Drazen JM. COX-2 inhibitors-a lesson in unexpected problems. N Engl

J Med. 2005;352:1131-1132.

Address correspondence to: Weng-Foung Huang, PhD, Institute of Health and Welfare Policy, National Yang-Ming University, 155 Li-Nong Street Section 2, Shih-Pai, Taipei, Taiwan 11221. E-mail: [email protected]