Nsaid Aki Harvard Ajm 2008

8/3/2019 Nsaid Aki Harvard Ajm 2008

1/7

CLINICAL RESEARCH STUDY

Nonselective and Cyclooxygenase-2-Selective NSAIDs and

Acute Kidney InjuryWolfgang C. Winkelmayer, MD, ScD, MPH,a,b Sushrut S. Waikar, MD, MPH,b Helen Mogun, MS,a

Daniel H. Solomon, MD, MPHa,c

aThe Division of Pharmacoepidemiology and Pharmacoeconomics, b Renal Division, cDivision of Rheumatology, Department of

Medicine, Brigham and Womens Hospital, Harvard Medical School, Boston, Mass.

ABSTRACT

OBJECTIVE: The association between nonsteroidal anti-inflammatory drugs (NSAIDs) and acute kidney

injury is well established, but it is less clear whether this risk is focused with specific agents. We undertooka large pharmacoepidemiologic analysis of the risk of acute kidney injury among older adults using

nonselective NSAIDs or cyclooxygenase (COX)-2 inhibitors.

METHODS: Medicare beneficiaries from 2 large states with drug benefit were eligible for study. Patients

were included if they filled a prescription for a nonselective NSAID or COX-2 inhibitor after more than

6 months without any such prescriptions and without a previous diagnosis of acute kidney injury. Incident

acute kidney injury was ascertained from hospitalization claims within 45 days of initiating nonselective

NSAID or COX-2 inhibitor therapy. Adjusted proportional hazards models estimated the relative risk of

acute kidney injury associated with each agent compared with celecoxib.

RESULTS: We included 183,446 patients whose mean age was 78 years; 80% were women. Acute kidney

injury was identified in 870 (0.47%) of nonselective NSAID or COX-2 inhibitor users. The agents with

significantly elevated risk compared with celecoxib were indomethacin (rate ratio [RR] 2.23; 95%

confidence interval [CI], 1.70-2.93), ibuprofen (RR 1.73; 95% CI, 1.36-2.19), and rofecoxib (RR

1.52; 95% CI, 1.26-1.83). These findings were robust in several subgroups.

CONCLUSION: Acute kidney injury requiring hospitalization is a relatively rare adverse event among older

adults after initiation of nonselective NSAIDs or COX-2 inhibitor treatment, observed in approximately 1

in 200 new users within 45 days. There seems to be a marked gradient of risk for acute kidney injury across

agents, specifically for indomethacin, ibuprofen, and rofecoxib.

2008 Elsevier Inc. All rights reserved. The American Journal of Medicine (2008) 121, 1092-1098

KEYWORDS: Acute renal failure; Adverse event; NSAID; Pharmacoepidemiology

Funding: No external funding.

Conflict of interest: Dr Winkelmayer is supported by a ScientistDevelopment Grant from the American Heart Association (0535232N), a

Norman S. Coplon Extramural Research Program Award from Satellite

Healthcare, Inc, and investigator-initiated grants from Amgen, Fresenius

Medical Care, and GlaxoSmithKline. He has participated, without receiv-

ing an honorarium, in advisory boards of Amgen, Roche, Genzyme, and

Fresenius. He has no personal financial relationships with any pharmaceu-

tical company. Dr Solomon has served as the Principal Investigator on

research grants in the past 5 years from Merck, Pfizer, and Savient. None

of these are currently active. He has received salary support from a research

grant from GSK in the past 3 years. This is not currently active. He serves

as an unpaid member of the Executive Committee for a Pfizer sponsored

clinical trial regarding NSAIDs. He also has served as an unpaid member

of Advisory Boards for Amgen and Abbott. He has no personal financialrelationships with any pharmaceutical company. Dr Waikar and Helen

Mogun have no potential conflicts of interest to report.

Authorship: All authors had access to the data. Drs Winkelmayer and

Solomon collaborated closely on designing the study, developing the an-

alytic protocols, interpreting the statistical output, and writing the article.

Dr Waikar helped interpret the results and reviewed and refined the article

draft, and Helen Mogun conducted the statistical analyses.

Requests for reprints should be addressed to Wolfgang C. Winkel-

mayer, MD, ScD, Division of Pharmacoepidemiology and Pharmacoeco-

nomics, Brigham and Womens Hospital, 1620 Tremont Street, Suite 3030,

Boston, MA 02120.

E-mail address: [email protected]

0002-9343/$ -see front matter 2008 Elsevier Inc. All rights reserved.

doi:10.1016/j.amjmed.2008.06.035
mailto:[email protected]:[email protected]


2/7

Nonsteroidal anti-inflammatory drugs (NSAIDs) are among

the most widely used medications in the United States,1

several of them being freely available over the counter.

Because barriers to self-medication are low, adequate fol-

low-up after initiating such treatment is often missing. Sev-

eral important side effects of

NSAIDs have been well docu-

mented, including gastroduodenalulcer disease and gastrointestinal

bleeding. To avoid these impor-

tant side effects, a new type of

NSAIDs was developed and intro-

duced into clinical practice: agents

with predominant activity on the

COX-2-enzyme, the COX-2 inhib-

itors.

Several previous publications

have linked NSAID exposure to an

increased risk of development of

chronic kidney disease,2,3 whileother reports have found an in-

creased risk of acute kidney in-

jury,4-6 but there is evidence of

heterogeneity in these risks among

agents in this class. Indomethacin

has previously been incriminated

to cause acute kidney injury, with

ibuprofen, piroxicam, and feno-

profen also exhibiting such risks.4 Most of the studies of

NSAIDs and acute kidney injury, however, were conducted

during the preCOX-2 inhibitor era.4,5,7 A recent well-

conducted population-based study has investigated the as-sociation between NSAIDs, including nonspecific NSAIDs

and COX-2 inhibitors, and acute kidney injury. The inves-

tigators found that several NSAIDs, including rofecoxib and

celecoxib, were associated with an increased acute kidney

injury risk compared with nonuser controls.8 It seemed that

the risk for acute kidney injury was smaller among cele-

coxib users compared with rofecoxib and most other

NSAIDs, but formal tests were nonsignificant. To further

clarify these relationships, we investigated short-term dif-

ferences in the risk of acute kidney injury among new users

of NSAIDs, with special focus on differences among

COX-2 inhibitors.

MATERIALS AND METHODS

Data Source and Study PopulationFor this study, we examined patients enrolled in Medicare

and either the New Jersey Pharmaceutical Assistance for the

Aged and Disabled program or the Pennsylvania Pharma-

ceutical Assistance Contract for the Elderly program be-

tween 1999 and 2004. These programs provide comprehen-

sive prescription drug coverage for eligible patients in New

Jersey and Pennsylvania with minimal copayments. Eligi-

bility is income-based and includes those individuals agedmore than 65 years who do not qualify for Medicaid but still

have limited income. Neither program has any restrictions

or prior authorization programs in place for any of the

NSAIDs studied, but no data on over-the-counter use are

included. The claims data from these pharmaceutical bene-

fits programs were merged with Medicare Part A and B

claims for these patients. We iden-

tified all patients who filled a pre-

scription for any NSAID (index pre-scription), nonselective or COX-2-

selective, after not having received

a prescribed NSAID in the previous

6 months. We then retained only

those who had been active users of

the program as indicated by at least

1 claim both 6 months and more

than 6 months before the filling date

of the NSAID prescription. Patients

younger than 65 years on the index

date also were eliminated, as were

patients who had been diagnosedwith acute renal failure or whose

claims indicated any maintenance

dialysis or kidney transplantation

care before the index date.

Main ExposuresFrom the index prescription, we

categorized patients as users of a

given NSAID, including celecoxib, rofecoxib, valdecoxib,

diclofenac, ibuprofen, indomethacin, meloxicam, nabum-

etone, naproxen, oxaprozin, sulindac, and other NSAIDs.

Prescription date, prescribed daily dose, and number of days

of treatment supplied were noted for each NSAID prescrip-

tion. We excluded patients who received more than 1

NSAID on the index date. Patients were followed up until

the earliest of the following events: filling a prescription for

an NSAID different than the one on the index prescription,

45 days after the index date, death, and end of database.

Other Patient CharacteristicsFrom enrollment files, we defined each patients age on

index date, gender, and race (white, black/other). From

claims spanning the 6 months before the index prescription,we ascertained several patient characteristics:9 We defined

several comorbidities, including congestive heart failure,

hypertension, myocardial infarction, peripheral artery dis-

ease, diabetes, gout, liver disease, malignancy, and renal

disease. The use of other relevant medications, such as

angiotensin-converting enzyme inhibitors (ACEIs) or an-

giotensin receptor blockers (ARBs), alpha-receptor block-

ers, beta-receptor blockers, calcium channel-blockers, di-

uretics, and loop diuretics, also was ascertained. We further

defined indicators of previous health care use, including the

number of hospital days, physician visits, any nursing home

stay, and total number of different medications in the pre-vious 6 months, as well as the calendar year of index

CLINICAL SIGNIFICANCE

Approximately 1 in 200 patients agedmore than 65 years will develop acutekidney injury within 45 days of newly ini-tiated NSAID therapy.

Patients initiating treatment with indo-methacin, ibuprofen, and rofecoxib hadhigher rates of acute kidney injury fromcelecoxib.

In light of newer COX-2 inhibitorsapproaching the market, clinicians needto be aware of the heterogeneity in acutekidney injury risk among COX-2 inhibitorsand demand pertinent safety data forthese new drugs.

1093Winkelmayer et al NSAIDs, COX-2-Inhibitors, and Acute Kidney Injury


3/7

prescription. Immediately preceding the index date (within

30 days), we also ascertained whether any tests or proce-

dures requiring administration of contrast media were con-

ducted or whether a patient underwent any surgery.

OutcomesThe outcome of this study was acute kidney injury as

ascertained from the presence of a diagnosis code indicating

acute renal failure (International Classification of Diseases,

release 9, code 584.5, 584.6, 584.7, 584.8, or 584.9). Pri-

mary outcome was the presence of such a code in conjunc-

tion with a hospitalization. This approach has been validated

and found to be highly specific (98%).10 The secondary out-

come was the presence of such a diagnosis code in either an

outpatient or inpatient claim.

Statistical AnalysesPatient characteristics were assessed across drug exposure

categories. We calculated crude incidence rates by dividingthe observed number of outcomes by the sum of all person

time within each exposure group. Unadjusted and fully

adjusted Cox proportional hazards models were then built to

estimate the relative risks of acute kidney injury; these

models were stratified for year and state.11 We chose cele-

coxib as the reference group, because it was the most fre-

quently used NSAID in this cohort and allowed for direct

comparison with the other COX-2 inhibitors, our primary

study aim. We estimated the hazard ratios [expressed as rate

ratios (RRs) in this article] and the corresponding 95%

confidence intervals (CIs). We included all baseline vari-

ables regardless of any significance threshold. Main analy-ses were unadjusted for multiple comparisons, as has been

suggested for exploratory studies;12 we also investigated

which of the apparent associations would be rendered

insignificant after Bonferroni adjustment for multiple

testing.13

For NSAIDs exhibiting an apparent association with

acute kidney injury, we defined baseline daily dose to assess

whether a dose-response relationship existed. We also con-

ducted subgroup analyses for important strata, such as by

previous use of ACEIs or ARBs, loop diuretics, diagnosis of

gout or use of any anti-gout medications, previous diagnosis

of chronic kidney disease, diabetes, congestive heart failure,or myocardial infarction.

We used SAS for Windows software (release 8.2; SAS

Institute Inc, Cary, NC) for all statistical analyses. This

study was approved by the institutional review board of

Brigham and Womens Hospital.

RESULTS

Cohort Selection and CharacterizationFrom the selection algorithm, we identified 183,446 study

patients aged more than 65 years who filled an NSAID

prescription after not having filled an NSAID prescriptionfor at least 6 months. The mean age was 78 years, and

80.3% were women. Celecoxib was the most frequently

used medication, with one third of patients (33.0%) using

this drug. Other frequently used NSAIDs were rofecoxib

(19.3%), ibuprofen (9.7%), naproxen (8.4%), and ketorolac

(7.0%). Table 1 provides the detailed characteristics of these

patients, stratified by NSAID used.

Risk of Acute Kidney InjuryBy using the more restrictive algorithm of ascertaining

acute kidney injury from a hospital claim, we identified 870

patients (0.47%) who experienced such an event during 45

days of follow-up. Including outpatient claims with a diag-

nosis of acute kidney injury, the secondary end point, the

number of affected patients increased only slightly, to 962

(0.52%). The crude incidence rate for the primary outcome

varied considerably among agents, from 0.02 per person-

year with meloxicam to 0.11 per person-year with indo-

methacin (Table 2). Celecoxib, the reference group, had an

incidence rate of 0.03 acute kidney injury events per person-

year of observation. In univariate analyses, there was clearevidence of heterogeneity in acute kidney injury rate among

the medications studied (P .001 from a global Wald test).

In univariate analyses, rofecoxib, ibuprofen, and indometh-

acin seemed to have a greater rate of acute kidney injury

compared with celecoxib. Multivariate adjustment con-

firmed the presence of these associations. Rofecoxib was

associated with a 52% increased rate of acute kidney injury

compared with celecoxib (95% CI, 26%-83%), whereas

ibuprofen had a relative rate of 1.73 (95% CI, 1.36-2.19)

and indomethacin had a relative rate of 2.23 (95% CI,

1.70-2.93; Table 3). These associations remained statisti-

cally significant after adjustment for multiple comparisons,whereas the apparent association of naproxen with acute

kidney injury (RR 1.37; 95% CI, 1.04-1.81) was no

longer significant after such adjustment. These results were

essentially unchanged in analyses of the combined outcome

of acute kidney injury in outpatient or inpatient encounters

(Table 3). When investigating the role of prescribed daily

dose on the rate of acute kidney injury, we found no dif-

ference in the rate of acute kidney injury among ibuprofen

or indomethacin users. Patients who received a higher daily

dose of rofecoxib, however, seemed to be at substantially

greater risk compared with patients with a lower prescribed

daily dose: Compared with those receiving a daily dose of12.5 mg or less (reference group for this analysis), patients

receiving more than 25 mg had a 77% increased risk of

acute kidney injury (RR 1.78; 95% CI, 1.14-2.78). The

risk of those receiving an intermediate dose (12.5-25

mg/d), the risk did not significantly differ from the lowest

daily dose group (RR 1.13; 95% CI, 0.83-1.57). Cele-

coxib, by contrast, did not exhibit an association between

the dose and the risk of acute kidney injury.

Sensitivity analyses that censored patients 15 days after

the days supplied in the index prescription expired yielded

essentially identical results. We further examined several

substrata, such as those defined by previous use of ACEIs orARBs, loop diuretics, diagnosed kidney disease, congestive

1094 The American Journal of Medicine, Vol 121, No 12, December 2008


4/7

Table 1 Characteristics of Study Patients by Specific Nonsteroidal Anti-inflammatory Medication

COX-2 Inhibitors Nonselective NSAIDs

Substance Celecoxib Rofecoxib Valdecoxib Diclofenac Ibuprofen Indomethacin Ketorolac Meloxicam Nabumetone NaproN in study 60,568 35,368 9,802 6,450 17,795 6,027 12,854 3,739 7,125 15,452

Percent 33.0 19.3 5.3 3.5 9.7 3.3 7.0 2.0 3.9 8

Age (y, mean) 78.4 78.2 76.6 76.7 73.4 77.0 78.5 76.5 76.4 74

Women 83.1 82.9 81.7 81.1 73.6 61.6 79.8 82.8 82.3 76

White race 89.1 90.0 89.4 86.7 80.0 85.1 90.2 88.0 85.3 83

Diabetes 13.9 13.7 15.1 14.1 14.2 15.6 18.0 15.1 13.1 14

Hypertension 57.1 56.2 58.7 54.3 52.2 60.9 58.7 59.0 54.3 52

Gout 4.1 3.9 3.7 4.0 3.1 24.5 3.8 3.4 3.1 3

Renal disease 1.9 2.0 2.3 1.4 1.8 4.8 3.2 2.0 1.3 1

MI 4.5 4.5 4.5 3.7 4.0 6.4 4.9 3.3 3.4 3

CHF 4.7 4.7 3.1 2.8 3.8 6.9 3.9 2.5 3.1 2

PVD 9.4 9.4 10.2 7.2 7.6 8.7 9.3 9.0 7.9 7

Liver disease 0.4 0.4 0.4 0.3 0.6 0.5 0.4 0.2 0.4 0Malignancy 2.2 2.4 1.7 1.5 2.7 2.5 1.8 1.2 1.4 2

ACEIs 28.5 28.4 28.6 27.3 27.7 35.1 30.6 29.7 26.4 28

ARBs 12.4 12.7 19.3 10.9 10.3 12.1 13.6 19.4 10.8 11

-agonists 6.1 5.7 4.8 6.4 6.3 9.0 6.4 5.0 6.3 6

-blockers 28.9 29.7 33.7 26.7 26.0 37.1 31.6 31.3 26.7 26

CCBs 33.9 32.9 31.4 31.6 30.6 33.0 33.8 32.8 32.6 29

Vasodilators 0.6 0.6 0.5 0.6 0.7 1.4 0.8 0.5 0.5 0

Diuretics 26.4 26.1 30.0 25.5 23.6 33.3 27.1 30.1 26.4 24

Loop-diuretics 22.3 21.7 20.4 18.0 17.5 31.3 21.6 19.2 16.9 16

Medications 7.3 7.3 7.5 6.8 7.1 7.3 7.6 7.4 6.7 6

Contrast 2.0 2.2 1.9 1.3 2.7 2.7 1.5 1.0 1.4 2

Surgery 0.5 0.6 0.4 0.3 0.7 1.2 0.3 0.2 0.2 0

MD visits 5.0 5.0 5.6 4.8 4.6 4.9 5.5 5.5 4.8 4Hospital days 1.3 1.3 0.4 0.9 1.2 1.1 0.7 0.4 1.0 0

Nursing home 5.1 5.2 3.7 2.7 3.9 2.8 2.4 2.6 2.9 2

COX cyclooxygenase; NSAID nonsteroidal anti-inflammatory drug; MI myocardial infarction; CHF congestive heart failure; PVD pulmonary vascular

inhibitor; ARB angiotensin receptor blockers; CCB calcium channel blocker.


5/7

heart failure, or gout. None of these revealed a marked

deviation from the overall findings of this study.

DISCUSSIONIn a large, population-based cohort of elderly patients, we

studied the risk of acute kidney injury among new users of

nonselective and COX-2-selective NSAIDs. We found thatapproximately 1 in 200 patients were diagnosed with acute

kidney injury during a hospitalization within 45 days of

initiating such treatment. This risk is most likely an under-

estimate of all kidney injury after NSAID initiation, because

only patients who were hospitalized and diagnosed with

acute kidney injury were captured. We also found that the

risk of acute kidney injury varied considerably among

agents. Although most NSAIDs had similar risks compared

with celecoxib users, some agents had a higher risk, such as

rofecoxib, ibuprofen, and indomethacin users (by 50%,

75%, and 100%, respectively).

Although these findings were robust in several subgroupanalyses, it is important to discuss these findings in light of

several limitations of our study. Several NSAIDs were

available over the counter without prescription, and patients

may have used these drugs before their index date. Although

this is possible, misclassification typically leads to a bias

toward the null. In addition, previous use of certain over-

the-counter NSAIDs would bias the studies away from an

association, because patients susceptible to developing

acute kidney injury from NSAID exposure would have been

more likely to be diagnosed with acute kidney injury and

thus excluded from the cohort. Second, we ascertained acute

kidney injury from diagnosis codes in provider billingclaims rather than from serial laboratory assessments of

kidney function. In previous work, we found that billing

codes indicating acute kidney injury in hospital claims had

high specificity (98%) but low sensitivity (35%), and that

sensitivity was lower for less severe cases of acute kidney

injury.10 It is therefore likely that we have underestimated

the incidence of acute kidney injury after NSAID adminis-

tration. It is possible that kidney function parameters were

already elevated before NSAID exposure or that patients

were incorrectly coded with or without a diagnosis of acute

kidney injury. In terms of classification of outcomes, our

previous work on billing codes for acute kidney injuryestimated a positive predictive value of 48% and negative

predictive value of 98%. To introduce bias to our findings,

the accuracy of acute kidney injury ascertainment would

need to vary across NSAIDs; we have no reason to believe

that this is the case. It is noteworthy that baseline hospital

days and physician visits were comparable across users of

celecoxib, rofecoxib, ibuprofen, and indomethacin, the

agents among whom differences in acute kidney injury risk

were found (Table 1). Differential ascertainment of acute

kidney injury because of physician encounters is therefore

unlikely to account for our findings. Unmeasured confound-

ing, especially by indication, remains a possibility. To theextent that International Classification of Diseases, 9th Re-T

able

2

Eventand45-DayFollow-updetails,

bySpecificNonsteroidalAnti-inflammatoryMedication(PrimaryEndPointofHospital-diagnosedAcuteKidneyI

njury)

COX-2inhibitors

NonselectiveNSAIDs

Subst

ance

Celecoxib

Rofecoxib

Valdecoxib

Diclofenac

Ibuprofen

Indo

methacin

Ketorolac

Meloxicam

Nabumetone

Naproxen

Oxaprozin

Su

lindac

OtherNSAIDs

Ninstudy

60,5

68

35,3

68

9,8

02

6,4

50

17,7

95

6,0

27

12,8

54

3,7

39

7,1

25

15,4

52

2,6

45

1,361

4,2

60

Perso

n-years

14,8

52

7,9

41

2,1

34

1,0

44

1,6

96

510

1,2

71

764

1,1

05

1,8

25

365

199

568

Events

433

358

67

28

89

57

34

15

26

63

12

8

12

Eventrate

0.0

29

0.0

45

0.0

31

0.0

27

0.0

52

0.1

12

0.0

27

0.0

20

0.0

24

0.0

35

0.0

33

0.0

40

0.0

21

C

OX

cyclooxygenase;NSAID

nonsteroidalanti-inflammatorydrug.



6/7

vision-Clinical Modification codes capture comorbidity data

accurately, we controlled for major medical conditions that

could confound the association between NSAIDs and the

risk of acute kidney injury.14,15 More general limitations of

retrospective claims database research include unavailabil-

ity of vital and laboratory parameters and of quantitative

information on comorbid conditions. Availability of such

information would make it less likely for any associations to

be residually confounded. Only a randomized trial could

remove any residual bias to the greatest extent possible.Further, although patients filled a prescription, they may not

have taken the medication received. Again, this misclassi-

fication would introduce bias toward a null finding. To

introduce bias away from the null, such behavior would

need to vary systematically across NSAIDs. Finally, we

chose to compare the risks of acute kidney injury among

new initiators of NSAID therapy but did not compare their

risks with patients who did not receive any NSAIDs. By

restricting the study to incident users of the study drugs, we

avoided the daunting task of controlling for confounding by

indication between users and nonusers.

The strengths of our study include the availability of alarge population-based cohort of elderly individuals who

had generous prescription drug coverage from state benefit

plans. We generated an inception cohort of new NSAID

users, which is a preferred way to study drug effects in

nonrandomized studies. We restricted the timeframe to 45

days after filling the index prescription to avoid exposure

misclassification and to study acute renal outcomes rather

than possible influences on the risk of chronic kidney dis-

ease or chronic changes in kidney function. Despite the

short time frame, we ascertained a large number of acute

renal failure outcomes.

Previous studies examining the risk of acute renal failurewith different NSAIDs are few. The Celecoxib Long-term

Arthritis Safety Study, a randomized controlled trial with

celecoxib versus 2 nonspecific NSAIDs (diclofenac and

ibuprofen),16 found a similar decline of renal function in

celecoxib versus ibuprofen users, with diclofenac users ex-

periencing a slightly greater decline compared with cele-

coxib. No assessment of acute kidney injury was conducted

in this study.

Zhang and colleagues17 conducted a study-level meta-

analysis of 114 trial reports of different COX-2 inhibitors

versus their respective trial controls. The renal outcomesstudied included renal dysfunction, hypertension, and pe-

ripheral edema. Their findings suggested different effects on

renal outcomes for different COX-2 inhibitors. Although

they described an increased risk of renal outcomes in pa-

tients randomized to rofecoxib versus the respective control

arms, a decreased risk was found for patients in celecoxib

groups compared with their controls. Aside from the meth-

odological limitations of such meta-analyses combining dif-

ferent exposure and control regimens, considerable hetero-

geneity was present in the way these renal outcomes were

defined in each study.

Few observational studies of the differential risk of acutekidney injury among nonspecific NSAIDs have been con-

ducted. One study compared nonspecific NSAID users with

non-NSAID users and found a dose-dependent risk of acute

kidney injury among nonspecific NSAID users compared

with nonusers.5 Another study confirmed this observation

and further found that the increased risk resided among

patients who received ibuprofen, piroxicam, or indometha-

cin, but not among users of other nonspecific NSAIDs.4 Yet

another study investigated 4 individual NSAIDs (diclofe-

nac, ibuprofen, meloxicam, and naproxen) and compared

their risks with those of patients without NSAID use. Al-

though the estimates of effect seemed to be compatible withheterogeneity of acute kidney injury risk across these

Table 3 Associations Between Individual Nonsteroidal Anti-inflammatory Medication and Acute Kidney Injury

End Point AKI Diagnosis in Hospital

(870 Outcomes) End Point any AKI Diagnosis (962 Outcomes)

Unadjusted Adjusted Unadjusted Adjusted

Celecoxib 1.0 1.0 1.0 (Referent) 1.0 (Referent)

Rofecoxib 1.50 (1.25-1.81) 1.52 (1.26-1.83) 1.45 (1.21-1.73) 1.46 (1.22-1.74)Valdecoxib 1.02 (0.74-1.40) 1.09 (0.79-1.51) 1.01 (0.74-1.38) 1.06 (0.77-1.44)

Diclofenac 0.88 (0.57-1.39) 1.08 (0.69-1.70) 0.92 (0.60-1.41) 1.06 (0.79-1.62)

Ibuprofen 1.42 (1.12-1.80) 1.73 (1.36-2.19) 1.57 (1.25-1.96) 1.69 (1.35-2.11)

Indomethacin 3.12 (2.40-4.04) 2.23 (1.70-2.93) 2.84 (2.21-3.65) 2.15 (1.66-2.78)

Ketorolac 0.83 (0.60-1.14) 0.79 (0.57-1.10) 0.86 (0.64-1.16) 0.83 (0.62-1.12)

Meloxicam 0.83 (0.60-1.14) 0.89 (0.51-1.57) 0.93 (0.56-1.55) 0.99 (0.59-1.65)

Nabumetone 0.71 (0.44-1.15) 0.89 (0.55-1.44) 0.68 (0.42-1.10) 0.80 (0.49-1.28)

Naproxen 1.07 (0.82-1.41) 1.37 (1.04-1.81)* 1.13 (0.87-1.48) 1.30 (1.00-1.69)*

Oxaprozin 1.22 (0.66-2.23) 1.61 (0.88-2.95) 1.41 (0.81-2.47) 1.71 (0.98-2.99)

Sulindac 1.49 (0.74-3.02) 1.31 (0.65-2.66) 2.20 (1.26-3.84) 1.85 (1.06-3.24)*

Other NSAID 0.75 (0.42-1.34) 1.00 (0.56-1.79) 0.81 (0.46-1.41) 0.96 (0.55-1.67)

AKI acute kidney injury; NSAID nonsteroidal anti-inflammatory drug.

*Adjustment for multiple comparisons rendered these associations nonsignificant.

1097Winkelmayer et al NSAIDs, COX-2-Inhibitors, and Acute Kidney Injury


7/7

NSAIDs, no formal testing was conducted.6 The total num-

ber of acute kidney injury cases (N 103) in that study

limited any meaningful multivariate analyses and yielded

rather low statistical power. All these studies were con-

ducted before the availability of COX-2 inhibitors. Most

relevant to our work is a recent case-control study from

Quebec province to investigate the risk of acute kidney

injury from NSAIDs, both non-specific NSAIDs andCOX-2 inhibitors.8 Similar to our study, follow-up was

short (30 days) and acute kidney injury also was ascertained

from diagnosis codes in hospital claims. In their study, only

meloxicam, naproxen, celecoxib, and rofecoxib users were

studied in separate exposure groups; all other NSAIDs were

aggregated into a single exposure group. In contrast with

our work, they used non-NSAID users as the reference

group. After multivariate adjustment, they found an in-

creased risk of similar magnitude for users of rofecoxib

(RR 2.3), naproxen (RR 2.4), and other nonspecific

NSAIDs (RR 2.3) compared with nonusers of NSAIDs,

whereas celecoxib was associated with an apparent butsmaller risk (RR 1.5). A formal test for interaction be-

tween rofecoxib and celecoxib was nonsignificant, but cer-

tainly underpowered.8 Our study, although different in de-

sign, formally confirms a higher acute kidney injury risk

with rofecoxib compared with celecoxib.

CONCLUSIONSFrom our data and that of others, 2 important lessons can be

learned. First, the assumption of a homogenous effect on the

risk of acute kidney injury among COX-2 inhibitors does

not seem to be supported by the available evidence. Ab-sence of a homogenous effect among COX-2 inhibitors has

been postulated by others based on findings from basic

research,18 and our findings support this hypothesis using

patient-level data. Although celecoxib seems relatively safe

with regard to renal outcomes and did not exhibit a dose-

response with acute kidney injury within strata of daily

dose, exposure to rofecoxib in general, and to higher doses

(25 mg/d) in particular, were associated with an increased

risk of acute kidney injury. This information is important

going forward, because newer COX-2 inhibitors are cur-

rently being investigated for clinical use. Second, we con-

firm previous reports of substantially increased risks ofacute renal impairment with indomethacin and ibuprofen.

Given the widespread use of these agents, prescribers need

to weigh the potentially increased renal risks from these

agents and the specific benefits from these drugs. It seems

that preferred use of other NSAIDs, including celecoxib,

might help avoid some undesired renal outcomes of such

analgesic therapy.

References1. Curhan GC, Bullock AJ, Hankinson SE, et al. Frequency of use of

acetaminophen, nonsteroidal anti-inflammatory drugs, and aspirin in

US women. Pharmacoepidemiol Drug Saf. 2002;11:687-693.

2. Sturmer T, Erb A, Keller F, et al. Determinants of impaired renal

function with use of nonsteroidal anti-inflammatory drugs: the

importance of half-life and other medications. Am J Med. 2001;

111:521-527.

3. Curhan GC, Knight EL, Rosner B, et al. Lifetime nonnarcotic anal-

gesic use and decline in renal function in women. Arch Intern Med.

2004;164:1519-1524.

4. Griffin MR, Yared A, Ray WA. Nonsteroidal antiinflammatory drugs

and acute renal failure in elderly persons. Am J Epidemiol. 2000;151:

488-496.

5. Perez Gutthann S, Garcia Rodriguez LA, Raiford DS, et al. Nonste-

roidal anti-inflammatory drugs and the risk of hospitalization for acute

renal failure. Arch Intern Med. 1996;156:2433-2439.

6. Huerta C, Castellsague J, Varas-Lorenzo C, Garcia Rodriguez LA.

Nonsteroidal anti-inflammatory drugs and risk of ARF in the general

population. Am J Kidney Dis. 2005;45:531-539.

7. Evans JM, McGregor E, McMahon AD, et al. Non-steroidal anti-

inflammatory drugs and hospitalization for acute renal failure. QJM.

1995;88:551-557.

8. Schneider V, Levesque LE, Zhang B, et al. Association of selective

and conventional nonsteroidal antiinflammatory drugs with acute renal

failure: a population-based, nested case-control analysis. Am J Epide-

miol. 2006;164:881-889.

9. Solomon DH, Avorn J, Sturmer T, et al. Cardiovascular outcomes in

new users of coxibs and nonsteroidal antiinflammatory drugs: high-

risk subgroups and time course of risk. Arthritis Rheum. 2006;54:

1378-1389.

10. Waikar SS, Wald R, Chertow GM, et al. Validity of International

Classification of Diseases, Ninth Revision, Clinical Modification

Codes for Acute Renal Failure. J Am Soc Nephrol. 2006;17:1688-

1694.

11. Localio AR, Berlin JA, Ten Have TR, Kimmel SE. Adjustments for

center in multicenter studies: an overview. Ann Intern Med. 2001;135:

112-123.

12. Rothman KJ. No adjustments are needed for multiple comparisons.

Epidemiology. 1990;1:43-46.

13. Lagakos SW. The challenge of subgroup analysesreporting without

distorting. N Engl J Med. 2006;354:1667-1669.

14. Kieszak SM, Flanders WD, Kosinski AS, et al. A comparison of the

Charlson comorbidity index derived from medical record data and

administrative billing data. J Clin Epidemiol. 1999;52:137-142.

15. Humphries KH, Rankin JM, Carere RG, et al. Co-morbidity data in

outcomes research: are clinical data derived from administrative data-

bases a reliable alternative to chart review? J Clin Epidemiol. 2000;

53:343-349.

16. Whelton A, Lefkowith JL, West CR, Verburg KM. Cardiorenal effectsof celecoxib as compared with the nonsteroidal anti-inflammatory

drugs diclofenac and ibuprofen. Kidney Int. 2006;70:1495-1502.

17. Zhang J, Ding EL, Song Y. Adverse effects of cyclooxygenase 2

inhibitors on renal and arrhythmia events: meta-analysis of random-

ized trials. JAMA. 2006;296:1619-1632.

18. Hermann M, Luscher TF. Are there differences in the renal effects of

selective cyclo-oxygenase 2 inhibitors? Nat Clin Pract Nephrol. 2006;

2:174-175.


Nsaid Aki Harvard Ajm 2008

Documents

Transcript of Nsaid Aki Harvard Ajm 2008