Zhang et al.

BMC Nephrology (2017) 18:256

DOI 10.1186/s12882-017-0673-8

RESEARCH ARTICLE Open Access

Non-steroidal anti-inflammatory drug

induced acute kidney injury in the
community dwelling general population
and people with chronic kidney disease:
systematic review and meta-analysis
Xinyu Zhang1, Peter T Donnan1, Samira Bell2 and Bruce Guthrie1*

Abstract
Background: Non-steroidal anti-inflammatory drugs (NSAIDs) are a common cause of adverse drug events (ADEs), but
renal risks of NSAIDs are less well quantified than gastrointestinal and cardiac risks. This paper reports a systematic
review of published population-based observational studies examining the risk of acute kidney injury (AKI) associated
with NSAIDs in community-dwelling adults and those with pre-existing chronic kidney disease (CKD).
Methods: MEDLINE and EMBASE databases were searched until June 2016, and 3789 papers screened. Ten studies
reporting NSAID risk of AKI in the general population were included in random effects meta-analysis, of which five
additionally reported NSAID risk in people with CKD.
Results: In the general population, the pooled odds ratio (OR) of AKI for current NSAID exposure was 1.73 (95%CI 1.44
to 2.07), with somewhat higher risk observed in older people (OR 2.51, 95%CI 1.52 to 2.68). In people with CKD,
individual study OR of AKI due to current NSAID exposure ranged from 1.12 to 5.25, with pooled estimate OR 1.63
(95% CI 1.22 to 2.19).
Conclusions: No study reported baseline risk of AKI in different populations meaning absolute risks could not be
estimated, but baseline risk and therefore the absolute risk of NSAID exposure is likely to be higher in people with CKD
and older people. Large population based studies measuring AKI using current definitions and estimating the absolute
risk of harm are needed in order to better inform clinical decision making.
Keywords: Acute kidney injury, Chronic kidney disease, Non-steroidal anti-inflammatory drugs, Pharmacoepidemiology

Background and 21% in the United States (US) [2, 3]. NSAIDs are also
Non-steroidal anti-inflammatory drugs (NSAIDs) are commonly implicated in hospital admissions due to ADEs,
commonly prescribed in primary care for their analgesic, including those which are fatal [4], but gastrointestinal
antipyretic and anti-inflammatory effects. One in fifteen and cardiac toxicity are better quantified than renal tox-
US adults are actively prescribed NSAIDs at any one icity [5].
time [1], and in many countries low-dose preparations NSAIDs can reduce renal blood flow, cause tubular
are also available over-the-counter (OCT). Partly due to obstruction through crystal deposition, and induce direct
their widespread use, NSAIDs account for 25% of adverse cytotoxicity and cell-mediated immune injury mechanisms
drug events (ADEs) reported in the United Kingdom (UK) leading to the occurrence of acute kidney injury (AKI). An-
other symptom that is commonly caused by NSAIDs is
* Correspondence: [email protected] interstitial nephritis (AIN) which requires specialist review,
1
Division of Population Health Sciences, University of Dundee, The renal biopsy, high-dose corticosteroid and/or immunosup-
Mackenzie Building, Kirsty Semple Way, Dundee DD2 4BF, UK pressant treatments, and will normally be progression in
Full list of author information is available at the end of the article

© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Zhang et al. BMC Nephrology (2017) 18:256 Page 2 of 12

chronic kidney disease (CKD) [6]. Older age [7, 8] and the full text paper was retrieved for review. Authors of
underlying chronic kidney disease are also related to the original studies were not contacted. Study selection and
onset of AKI during NSAID use, with early studies show- quality assessment from full-text papers retrieved were
ing that the risk of deterioration in renal function increases performed independently by two reviewers (XZ and BG).
3–4 fold in patients with abnormal baseline renal function
compared to those with normal renal function [9]. Notably,
Study selection
NSAIDs are commonly prescribed to people with CKD,
The search strategy and data extraction were defined in
despite guidance to avoid them in this population. In US
a PICOS format (participants, intervention, comparison,
veterans in 2005, 15.4% of people with CKD were pre-
outcome and study design). Studies published in English
scribed traditional NSAIDs or COX-2 inhibitors [10], com-
were eligible for inclusion when they used observational
pared to 11.1% of people with CKD in the UK in 2012 [11],
methods to study adults in the community exposed to
and 15.9% of people with CKD in Australia in 2004–2006
NSAIDs and reported AKI as an outcome. Given the
[12]. Better quantification of risk in people with CKD is
historical lack of consensus on AKI definition, studies
therefore of particular clinical interest, as is whether
using a variety of definitions of AKI were included, with
NSAID risk varies by age and by COX-2 selectivity. In
AKI defined by International Classification of Diseases
terms of COX-2 selectivity, early studies suggested that
(ICD) 9 or 10 codes, or change in eGFR or creatinine clear-
COX-2-selective inhibitors caused fewer renal adverse
ance (CrCl) or serum creatinine (SCr). Similarly, varying
effects including reduction in glomerular filtration rate
definitions of CKD were allowed including estimated
(GFR), increased serum creatinine (SCr) and hyperten-
GFR < 60 ml/min (with or without standardization to body
sion [13–15]. Other studies have shown no significant
surface area), or based on ICD codes, or SCr > 122 μmol/L,
differences in renal risk between COX-2-selective in-
or structured patient interview. Traditional NSAIDs and
hibitors and nonselective NSAIDs [16, 17].
COX-2 inhibitors were included with the exception of low
There is little evidence about the risk of AKI associ-
dose aspirin (<300 mg per dose) [20]. Studies were ex-
ated with NSAID use in people with CKD available from
cluded if they were published in abstract only, included
randomised trials of NSAIDs because such trials rou-
children (age < 18 years old), only included post-operative
tinely exclude people with CKD and rarely report renal
patients or others receiving only one or two doses of
outcomes [18]. Under these circumstances, observational
NSAID as treatment (e.g. for renal colic or post-lithotripsy),
evidence provides the best guide to practice that exists.
or had end-stage renal disease defined as being on dia-
The aim of this study is to systematically review published
lysis or having received a renal transplant. Finally,
high-quality population-based observational studies to
meta-analyses, studies with <100 subjects, and studies
quantify the risk of AKI due to NSAIDs in the general
without a contemporaneous control group drawn from
population and in people with pre-existing CKD.
the same population were excluded.
Methods
Data sources and search strategy Data extraction and quality/validity assessment
MEDLINE and EMBASE were systematically searched Data were extracted into a standardised form and checked
from inception to June 21th 2016 using OVID from the for accuracy by a second reviewer. When data were re-
Knowledge Network using a predetermined list of key- ported in strata, the data were extracted as separate
words including NSAIDs, renal diseases and renal function subgroups. The following data were extracted for each
measurements modified from the search strategies used by included study: author, publication year, study design,
two related Cochrane reviews (see Additional file 1 for population (data source, sample size, location, age, gen-
search strategy) [18, 19]. Search results were restricted to der and underlying renal conditions), definition of AKI,
cross-sectional, cohort and case-control studies in the inclusion criteria, exclusion criteria, medication exposure
English language. The reference lists from all identified (type of NSAID), period and length of NSAID usage,
primary studies, review articles, Kidney Disease Improving number of people who were and were not exposed to
Global Outcomes (KDIGO) clinical practice guidelines for NSAIDs, as well as crude unadjusted and adjusted associa-
CKD and AKI and OpenSIGLE (unpublished literature tions between NSAID use and outcomes. The quality of
database) were manually checked to screen for additional the included studies was evaluated in three domains
relevant papers. using the validated Newcastle-Ottawa Quality Assess-
Citations were independently screened for eligibility by ment Scale for cohort and case-control studies [21], with
two reviewers based on title and abstract (XZ and SB or each item rated as either one star or missing (Table 1).
XZ and BG). If one or more authors deemed the study Disagreements were resolved by discussions with two
potentially relevant, or if there was any uncertainty authors (XZ and BG) and a third reviewer was involved
about eligibility based on title and abstract alone, then where required (PTD).
Zhang et al. BMC Nephrology (2017) 18:256 Page 3 of 12

Table 1 Quality assessment

a
If cases are first occurrence of outcome, then it must explicitly state that controls have no history of this outcome. If cases have new (not necessarily first)
occurrence of outcome, then controls with previous occurrences of outcome of interest will not be excluded
b
= same non-response rate for both groups reported ; NP = not reporting non-response rate is Not a Problem (since bias is less likely if the cases and controls
come from the same population and have outcomes and exposures ascertained in the same way)
c
CC case-control, N-CC nested case-control

Statistical analyses assumed that the OR was an accurate estimate of the

The outcome was the presence or not of AKI. In studies relative risk (RR) of AKI in NSAID users compared with
among the general population, adjusted odds ratio (OR) non-users.
with 95% confidence intervals (CI) for AKI with NSAID Statistical analyses were performed using Review
exposure were pooled using the generic inverse variance Manager 5.2 (Cochrane Collaboration, Oxford, United
method which assumed weights equivalent to inverse of Kingdom). Statistical significance was set at P < 0.05
variance of individual estimates [22, 23]. This was be- for all analyses. This systematic review was structured
cause adjusted ORs and CIs were mostly presented in in accordance with the Meta-analysis of observational
the primary studies while raw ORs and CIs were not. studies in epidemiology (MOOSE) statement (Additional
Moreover, pooled results will be more meaningful with file 2) [26].
the adjustment. When calculating the pooled result, in
order to be more conservative, the individual OR with the Results
most extreme of the lower or higher side of the CI for Study flow and characteristics
each study was used to estimate the variance. Additional Electronic searches retrieved 4629 citations, with 3789
subgroup meta-analyses were conducted to explore het- unique citations screened and four studies [27–30] iden-
erogeneity stratified according to pre-specified study-level tified from other sources (Fig. 1). After title and abstract
covariates namely age and COX-2 selectivity. screening 30 full-text studies were reviewed of which 10
In all but one case, primary studies which included studies published between 1990 and 2012 met the inclu-
analysis for the subgroup of people with CKD did not sion and quality criteria and were included (details for
provide adjusted estimates of association. Therefore crude excluding were recorded in Additional file 3). All 10
ORs and CIs from the raw data were calculated, and studies examined NSAID-associated AKI risk in the
pooled using Mantel-Haenszel method. general population with a total of 1,609,163 participants
A random-effects model was used for all analyses, and [6, 8, 9, 31–37]. Five of these studies also provided data
heterogeneity between studies assessed by the I2 statistic in the subset of people with CKD [31–33, 35, 37].
and the χ2 test for heterogeneity. I2 is the percentage of Eighty percent of studies were rated with seven or more
variance that is due to between-study variance and is an stars out of a possible nine on the Newcastle-Ottawa
indicator of consistency between studies. Values of 25– quality assessment scale (Table 1). Hence the quality
50%, 50–75% and >75% were considered evidence of of the included studies was considered to be medium
mild, moderate and marked heterogeneity, respectively to high.
[24]. Publication bias was not assessed because of the In all included studies, eligible cases with AKI were re-
extensive statistical heterogeneity found, since such het- cruited in a defined catchment area over a defined period
erogeneity in itself may lead to funnel plot asymmetry of time, the ascertainment of exposure was through secure
[25]. AKI is an uncommon adverse event and it was electronic records and the same method of ascertainment
Zhang et al. BMC Nephrology (2017) 18:256 Page 4 of 12

Fig. 1 Flow diagram of the identification process for eligible studies

was used among cases and controls. Eighty percent of the interviews. One study population comprised patients tak-
included studies had independent validation of cases while ing angiotensin-converting-enzyme inhibitor (ACE inhibi-
the remaining 20% relied on record linkage alone (ICD tors) and the others were of the general adult population
codes in database) with no reference to the primary [8]. Seventy percent of the studies focused on older partic-
record. One study [33] used hospital controls as their ipants (either only including older participants or where
cases were people with community exposure to NSAIDs the mean/ median age of participants was >65 years) while
admitted to hospital with AKI. Two studies [34, 37] used a gender proportions among included studies were diverse.
mixture of hospitalized and community controls. The Current or new use of NSAIDs as reported by the original
other included studies selected controls from community study authors was used as the exposure definition (70% of
or reported hospital and community controls separately in studies reported exposure to NSAIDs 0–90 days prior to
which case only comparisons with community controls the index day for the adverse event, other studies did not
were included. The history of outcome occurrence in specify). One study only examined ibuprofen [9], but most
cases and controls were adequate in 90% of the studies. examined exposure to a variety of NSAIDs. Half of the
Two included studies did not match for age and other studies used laboratory data to define the presence of AKI,
factors among cases and controls while one of them ad- whereas varying sets of hospitalisation discharge diagnosis
justed for a list of confounders including sex and use of ICD codes were used in the other half. Each of the studies
prescription of acetylsalicylic acid (ASA) in current use adjusted for a list of confounders. Newer studies tended to
of any NSAID. have a more thorough adjustment. The most common
Table 2 summarises the 10 included nested case-control confounders studies adjusted for general polulation are
and case-control studies (five each), of which five involved age, gender, comorbidity (such as hypertension, diabetes,
participants from North America, four from Europe and heart failure, and cardiovascular disease), concomitant drug
one from Australia. Nine studies used data extracted from use (such as diuretics, antibiotics, radio contrast ex-
routine electronic databases while the Australian study posure, and nephrotoxic drugs), and hospitalization
[33] combined electronic data with structured patient (Table 3).
Table 2 Characteristics of included studiesa
First Design; Data; Inclusion Criteria Exclusion Criteria Participants Mean Age % Male Definition of NSAID Definition of AKI Adjusted OR
Author, Country (yr) use (exposure)b (outcome) (95%CI) for general
Year population, plus
crude OR for CKD
pop if available
Bouvy N-CC; PHARMO >40 years, with ≥2 consecutive Hospitalisation with renal 144 cases Not 63.9 New/ start of ≥1 Hospitalisation ICD9 584 2.20 (1.10,4.50)
2003 [8] record linkage prescription for an ACEI problems before start an and 1189 reported cases; prescription in or 586
system; The ACEI controls (all 45.4 3 months before
Netherlands >40 years) controls hospital admission
Huerta N-CC; GPRD; UK 50–84 years, ≥2 years Cancer, renal disorder, 386,916 Not Not Supply for the Clinical diagnosis by a 3.23 (1.79,5.82)
2005 [34] enrolment with GP and ≥1 year cirrhosis, systemic individuals reported reported most recent specialist, and SCr
since first computerized connective tissue disease (all prescription lasted >1.7 mg/dl (150 μmol/L)
prescription >50 years) until 0–30 days or urea level >47.6 mg/dL
Zhang et al. BMC Nephrology (2017) 18:256

before index date (17.0 mmol/L)

Leonard CC; GPRD; UK Not reported History of kidney 27,982 68.6 cases; 49.7 Active orally Diagnostic codes 1.31 (1.25,1.37)
2012 [6] transplant, having cases and 66.9 cases; administered described in succeeding
outcome of interest during 1,323,850 controlsc 50.4 tNSAID therapy texts supplemented by
baseline period controls controls GP’s free-text
Murray CC; Regenstrief >18 years, received ibuprofen or Prescriptions of other 4790 cases Not 27.4 Patients received Patients with normal 1.05 (0.88,1.26)
1990 [9] Health Center; acetaminophen during NSAIDs, SCr < 0.3 mg/dL and 8205 reported cases first prescription of baseline values, SCr
US 11May1975- 29Sept1986, (30 μmol/L), BUN < 5 mg/ controls N/A ibuprofen between >1.2 mg/dL (110 μmol/L)
baseline and post-prescription dL (1.8 μmol/L) controls 11May1975- or BUN > 18 mg/dL
SCr and BUN results available 29Sept1986 (6.4 μmol/L);
Patients with elevated
baseline values, SCr or
BUN ≥10%increase
Perez CC; ≥1 NSAID prescription during Malignant neoplasm, CRF, 228,392 Not 45.5 Most recent ICD9 580.9, 581, 583.2, 4.10 (1.50,10.8)
Gutthann Saskatchewan study period in-hospital disease onset, members reported prescription filled 583.6–583.9, 584, 586,
1996 [36] health plan insufficient data, other sys- 0–30 days before 593.9
information; temic/ renal conditions index
Canada
First Design; Data; Inclusion Criteria Exclusion Criteria Study Mean Age % Male Definition of NSAID Definition of AKI Adjusted OR
Author, Country Sample (yr) use (exposure)b (outcome) (95%CI) for general
Year population, plus
crude OR for CKD
pop if available
Evans CC;MEMO; UK Resident in Tayside, Scotland Not reported 320 Not Not ≥1 oral NSAID ICD9 583.8, 584.5, 584.7– 2.20 (1.49,3.25);
1995 [31] registered with a Tayside GP in patients reported reported prescriptions 584.9 Crude OR for CKD
May 1990 and 1238 dispensed during population 3.04
community 90 day period prior (1.68,5.49)
controlsd to the index date
Griffin N-CC; Tennessee ≥65 years, enrolled in End-stage renal disease, 7145 Not 31 Nonaspirin, supply An admission SCr 1.58 (1.34,1.86);
2000 [32] Medicaid Medicaid ≥ 1 year hospital-acquired acute patients reported cases; of NSAIDs included ≥180 μmol/L (2 mg/dl) Crude OR for CKD
enrolment files; renal failure, incomplete and 10,000 (all 23 index date and ≥20% increase from population 1.80
US demographic data, remote controls ≥65 years) controls baseline or ≥20% decline (1.30, 2.50)
counties residents during hospitalization
ICD9 584, 586
Page 5 of 12
Table 2 Characteristics of included studiesa (Continued)
Schneider N-CC; Quebec >65 years, filled ≥1 NSAID Only use aspirin, renal 121,722 78.1 cases; 46.1 Dispensed NSAID 2.05 (1.61,2.60);
2006 [37] universal health prescription during 01Jan1999- replacement therapy, renal new NSAID 78.0 cases; 1–30 days Crude OR for CKD
care program 30June2002, NSAID prescription transplantation, 2 NSAIDs users controls 32.3 preceding the population 1.13
database; free ≥1 year before cohort entry at cohort entry controls index date with no (0.79, 1.62)
Canada previous
prescription
Lafrance N-CC; ≥1 NSAID prescription during History of renal 1,432,781 63 (half 97 Single NSAID Hospitalisation with AKI, 1.82 (1.68,1.98);
2009 [35] Department of 01Oct2000-30Sept2006, NSAID transplantation, new NSAID >65 years) dispensed AKIN definition Crude OR for CKD
Veterans Affairs prescription free 2 years before maintenance dialysis, or users day + 30 days population 1.36
(VA) health care cohort entry AKI before cohort entry tolerance period (1.30, 1.42)
system; US with no previous
prescription
Henry Matched CC; Admitted to study hospitals Unfit for interview 164 cases 76.6 cases; 55.5 Any NSAID use in Admitted to hospitals 1.80 (0.97,3.40);
Zhang et al. BMC Nephrology (2017) 18:256

1997 [33] John Hunter identified by hospital database and 189 75.1 cases; past month with SCr ≥0.15 mmol/L Crude OR for CKD
Hospital and controls controls 55.0 (excluding population
Newcastle controls prophylactic 5.25 (1.06,26.07)
Master Hospital; aspirin)
Australia
a
CC, case-control; N-CC, nested case-control; MEMO, Medicines Monitoring Unit’s record-linkage database; ICD9, International Classification of Disease version 9; AKIN, Acute Kidney Injury Network; ACEI, angiotensin-
converting-enzyme inhibitor; GPRD, General Practice Research Database; SCr, serum creatinine; tNSAID, traditional NSAID; BUN, blood urea nitrogen; CRF, chronic renal failure
b
Definition chosen by review authors
c
Age median (rather than mean)
d
There are hospital controls which were ignored
Page 6 of 12
Zhang et al. BMC Nephrology (2017) 18:256 Page 7 of 12

Table 3 Confounders that the included studies adjusted for

First Author, Year Confounders adjusted in general population
Bouvy 2003 [8] Age and gender, prior hospital admissions for congestive heart failure, diabetes and for concomitant use of diuretics,
low-dose aspirin, antibiotics, paracetamol (acetaminophen), epoetin, corticosteroids, opioids, digoxin, antigout drugs
and duration of use of ACE inhibitor
Huerta 2005 [34] Sex, age, calendar year, body mass index, HF, hypertension, diabetes, antihypertensive use, oral steroid use, NSAID use,
and consultant visits and hospitalizations in the previous year
Leonard 2012 [6] Hospitalized in prior 30 days, ever past anemia, ever past coronary disease, ever past heart failure/cardiomyopathy,
ever past disorders of stomach function, ever past arthropathies and related disorders, ever past pain, ever past
gastrointestinal drug use, ever past cardiovascular system drug use, ever past central nervous system drug use, ever
past infection-treating drug use, ever past endocrine system drug use, ever past nutrition and blood drug use, ever
past musculoskeletal and joint disease drug use, frusemide use in the prior 28 days, and kidney sensitizer drug
exposure in the prior 180 days
Murray 1990 [9] Age, gender, race, coronary artery disease, baseline systolic blood pressure, diuretic use,
Perez Gutthann 1996 [36] Age, sex, calendar year, cardiovascular risk indicator, recent hospitalization for disorders renal, exposure to NSAIDs,
prescription ASA, nephrotoxic drugs
Evans 1995 [31] Age, gender, could not find information in other covariates
Griffin 2000 [32] Age (65–74, 75–84, > = 85), gender, ethnicity, nursing home resident, recent hospitalization (within 30 days,
31–365 days, none in the past year), concomitant use of loop diuretic, thiazide, ACE inhibitor, and antibiotics
(within 30 days), prescription for allopurinol, cyclosporin, gold, sulfinpyrazone, or penidllamine, first prescription
for cimetidine in the past 60 days, or procedure code Indicating intravenous radio contrast within the past 30 days
Schneider 2006 [37] Age, gender, comorbidity (Hypertension, Diabetes, Heart failure, Cardiovascular disease, Atherosclerosis, Hyperlipidemia,
Respiratory disease, Gastrointestinal ulcer disease, Chronic renal failure, Acute renal failure, Renal disease, Renovascular
disease, Renal infection, Conditions secondary to renal impairment, Renal manifestation of systemic diseases, Systemic
disease and malignancy relevant to renal function), drug use (Oral anticoagulants, Oral corticosteroids, Psychotropic
drugs, Thyroid drugs. Current use of aspirin, Use of nephrotoxic drugs, Exposure to contrast media), comorbidity
measures (No. of different drugs, Chronic disease score, Charlson index, Health care utilization (>12 physician visits,
> = 1 nephrologist visits >1 hospitalization))
Lafrance 2009 [35] Age, gender, race, concurrent disease (Arrhythmia, Chronic kidney disease, Cardiovascular disease, Cancer, Chronic liver
disease, Chronic pulmonary disease, Congestive heart failure, Diabetes, Hyperlipidemia, Hypertension, Osteoarthritis,
Rheumatoid arthritis, Peptic ulcer/ GERD, PVD, Valvular disease), hospitalization (last 30 days, previous year), drug use
(ACEi or ARBs, Beta-blockers, Diuretics, Oral anticoagulants, Platelet aggregation drugs, Nephrotoxic drugs, Corticosteroids,
Radio contrast exposure), laboratory (Serum albumin)
Henry 1997 [33] Age, history of gout, heart disease and renal disease

Association between NSAID exposure and AKI in the (Fig. 2). The pooled crude OR was 1.63 (95% CI 1.22–
general population 2.19) and I2 statistic was 71% (P = 0.009). We noted
Ten studies that included a total of 1,609,163 partici- weaker associations with AKI in larger studies with more
pants were used to evaluate AKI risk among current precise estimates of risk, with the two studies reporting
NSAID users in the general population (Fig. 2). The ad- the largest risks being older, smaller and less precisely
justed OR for AKI was increased relative to non-users estimated.
and between individual studies ranged from 1.05 to 4.10,
with eight of the ten studies showing a statistically sig- Subgroup analyses
nificant association between NSAID exposure and AKI. To explore heterogeneity, we examined association with
Meta-analysis of adjusted odds ratios estimated the AKI in older users of NSAIDs (age > 50 years), stratified
pooled OR to be 1.73 (95%CI 1.44–2.07). Heterogeneity by COX-2 selectivity of the NSAIDs exposed to [38]
was substantial (I2 = 89%, P < 0.001), suggesting that the (none, <5-fold and ≥5-fold), and in older patients with
pooled estimate should be interpreted with caution. exposure to COX-2 selective NSAIDs (Figs. 3, 4 and 5).
Statistical heterogeneity remained in subgroup analyses
Association between NSAID exposure and AKI in people but it was modestly reduced, suggesting that subgroup
with CKD analyses provided more confidence in the pooled esti-
Two case-control and three nested-case-control studies mates, but interpretation of pooled estimates should still
which included 106,681 people with CKD reported data be cautious.
that could be used to evaluate AKI risk by current The pooled results of NSAIDs with different COX-2
NSAID exposure. The crude OR for AKI in NSAID selectivity for AKI are shown in Fig. 3. Similar to the
users compared with nonusers ranged from 1.12 to 5.25 general results, there were increased associations between
and was >1 and statistically significant in 4 of 5 studies AKI and exposure to NSAIDs with different COX-2
Zhang et al. BMC Nephrology (2017) 18:256 Page 8 of 12

Fig. 2 Pooled odds ratio and 95% confidence intervals for AKI in general population and people with CKD using NSAID vs not using

selectivity (no COX-2 selectivity 1.84, 95%CI 1.54–2.19, in order to try to explore and explain heterogeneity. Re-
COX-2 selectivity <5-fold 1.43, 95%CI 1.25–1.64 and sults were consistent with and similar to the main find-
COX-2 selectivity ≥5-fold 1.41, 95%CI 1.07–1.87). There ings. Older people who were prescribed NSAIDs had a
was a non-statistically significant trend that the higher the somewhat higher (2-fold) risk of developing AKI, but
COX-2 selectivity NSAIDs was, the lower the increased there was no strong evidence that greater COX-2 selectiv-
odds of AKI (subgroup difference test χ2 = 5.31, df = 2, ity was associated with lower AKI risk. NSAIDs with high
P = 0.07). COX-2 selectivity (≥5-fold) had a lower association with
Older people (>50 years) using NSAIDs had a some- AKI than NSAIDs with COX-2 selectivity <5-fold, and
what higher odds of AKI associated with NSAID expos- heterogeneity in the subgroups was reduced compared to
ure than the general population (2.01, 95%CI 1.52–2.68; the overall results consistent with some of the heterogen-
Fig. 4), although the confidence intervals overlap, and eity being due to differences in the age of the population
there was again significant heterogeneity (I2 = 62%). In studies and the type of NSAIDs examined. Five studies in-
older people who were using NSAIDs with COX-2 se- cluded individual NSAID usage in their analyses in which
lectivity, the odds ratio was 1.73 (95%CI 1.32–2.29; only Lafrance and Schneider compared dose effect in
Fig. 5) which is similar to the general population. Het- Rofecoxib, Celecoxib, Naproxen and Meloxicam (Lafrance
erogeneity was moderate (45%) in this subgroup. only) [9, 32, 34, 35, 37]. Dose response cannot be easily
stratified as higher dosage will be associated with a higher
Discussion risk of effect compared to lower dosage but the exposure
The results of the meta-analysis showed that current ex- window is not under control. However, other differences
posure to NSAIDs was associated with an approximately in population and in AKI definition were substantial and
1.5-fold increase in the odds of developing AKI in the likely accounts for much of the observed heterogeneity.
general population and in people with CKD. Since AKI Overall, all analyses showed a statistically significant, mod-
is a rare NSAIDs associated adverse outcome, odds ra- estly increased risk of AKI from exposure to NSAIDs, and
tios will approximate to relative risks. There was con- the magnitude of the increased risk of AKI was rather
siderable heterogeneity between studies, particularly in similar among all sub-groups with mostly overlapping
the general population group and so the pooled esti- confidence intervals.
mates should be interpreted with caution. The limited The study strengths include careful study selection
numbers of studies eligible for inclusion precluded and the use of a structured quality assessment tool to
meta-regression, so subgroup analyses were conducted ensure that only high quality studies were included [21].
Zhang et al. BMC Nephrology (2017) 18:256 Page 9 of 12

Fig. 3 Pooled odds ratio and 95% confidence interval for AKI in general population using NSAIDs with different COX-2 selectivity vs not using

The observed associations were consistent across sub- outcomes (the focus of this study) and exclude people
groups, but the study has several limitations. As with all with CKD (a key topic of interest) and other comorbidi-
systematic reviews, the findings depend on the quality of ties as well as older people and minority groups [18, 39].
the included studies. We chose to review and meta- However, observational studies are vulnerable to residual
analyse observational studies because an initial literature confounding by measured and unmeasured variables. An
search identified that trials of NSAIDs rarely report renal example is confounding by indication, which in this

Fig. 4 Pooled odds ratio and 95% confidence interval for AKI in elderly people using NSAIDs vs not using
Zhang et al. BMC Nephrology (2017) 18:256 Page 10 of 12

Fig. 5 Pooled odds ratio and 95% confidence interval for AKI in elderly people using NSAIDs with COX-2 selectivity vs not using

context is likely to occur if prescribers avoid NSAIDs in AKI in the general community-dwelling population and
people they perceive to be at higher risk of NSAID tox- people with CKD. A previous systematic review which
icity including AKI, which would lead to an underesti- was conducted in 2014 included five observational studies
mation of AKI risk if present. There were also large and reported risk of AKI by individual NSAID exposure,
differences between studies in the population examined finding a statistically significant elevated AKI risk among
and the way in which AKI was measured, both of which most of the traditional NSAIDs but did not achieve a stat-
likely contributed to the observed moderate to large het- istical significance for COX-2 inhibitors or traditional
erogeneity between studies. It is also important to recog- NSAIDs with higher COX-2 selectivity (meloxicam and
nise that the estimate of the risk of NSAIDs in the diclofenac) [40]. Our study used a more comprehensive
general population is adjusted for potential confounders, search and included additional studies, and found a
but the estimate in people with CKD is not because only similar estimate of pooled risk but statistically significantly
one study reported an adjusted estimate [33]. Other increased risk irrespective of COX-2 selectivity. Another
limitations include that we only included studies pub- systematic review specifically focused on myocardial, vascu-
lished in English, that there were a relatively small lar and renal risks of COX-2-selective meloxicam allowed a
number of studies suitable for inclusion which made broad definition of renal outcomes and it did not find a
meta-regression to explore heterogeneity unfeasible, significantly increased renal risk [41]. A third systematic
and that the rate of concomitant use of OTC NSAID review examined the risk of CKD progression associated
use could not be assessed in the populations studied. with NSAID use, finding that high (but not standard) dose
Seven of the included studies [6, 8, 31, 32, 34, 35, 37] NSAID use was associated with an increased risk of CKD
addressed that OCT NSAID use may have caused expos- progression [42].
ure misclassification. But all of the studies believed that
due to reasons such as financial incentives, the proportion Conclusion
of OCT NSAID users is expected to be small and nondif- AKI is an increasingly common global problem causing
ferential with respect to the NSAID categories the popula- significant morbidity and mortality and with large resource
tions studied. Consequently, it would bias the results implications. Exposure to NSAIDS and other nephrotoxic
toward the null. As most of the studies confirmed eligible drugs is an important cause of AKI, but the risk of these
cases then selected controls according to a certain propor- exposures is modified by susceptibilities such as increasing
tion, non-response rate was not given in the majority of age and the presence of CKD [43]. This study found that
the included studies. Since cases and controls of these the odds of developing AKI increased by over 50% in
studies were derived from the same databases, were exam- people who were exposed to NSAIDs in the general popu-
ined for the same exposure and followed up in the same lation and in people with CKD, and in older people the
way, then non-response rate is not considered to be a odds of developing AKI doubled. However, the absolute
problem and therefore it is unlikely there would be miss- risk of developing AKI also depends on the baseline risk of
ing data bias. We were unable to access publication bias AKI in the population exposed, which none of the included
because of the extensive statistical heterogeneity. For studies reported. Future studies should use internationally
many of the methodological qualities assessed there was agreed definitions of AKI [44] and estimate the absolute
an unclear risk of bias as studies did not provide explicit risk of AKI in different populations including older people
detail to make an informed judgement. and people with CKD to better inform clinical decision
To our knowledge, this study is the first meta-analysis making. There is evidence that feedback and more com-
to examine associations between NSAID exposure and plex interventions to reduce NSAID prescribing in people
Zhang et al. BMC Nephrology (2017) 18:256 Page 11 of 12

at high risk of renal adverse effects are effective [45, 46], Sciences and Amgen unrelated to this paper. The other authors declare no
and clinicians should seek to minimise NSAID expos- conflicts of interest.

ure in people particularly susceptible to AKI due to

age, CKD or because of the co-prescription of other Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in
nephrotoxic drugs [43, 47]. published maps and institutional affiliations.

Author details
Additional files 1
Division of Population Health Sciences, University of Dundee, The
Mackenzie Building, Kirsty Semple Way, Dundee DD2 4BF, UK. 2Renal Unit,
Additional file 1: Complete search strategy. The complete search strategy Ninewells Hospital, Dundee, UK.
for the systematic review for both Medline and Embase. (DOCX 22 kb)
Additional file 2: MOOSE Checklist. Essential items to report in meta-analysis Received: 17 March 2017 Accepted: 19 July 2017
of observational studies in epidemiology. (DOCX 26 kb)
Additional file 3: Detailed reasons for excluding full text. Detailed reasons
References
for 20 excluded papers at full text papers reviewing stage with reference.
1. Paulose-Ram R, Hirsch R, Dillon C, Losonczy K, Cooper M, Ostchega Y.
(DOCX 26 kb)
Prescription and non-prescription analgesic use among the US adult
population: results from the third National Health and nutrition examination
Abbreviations survey (NHANES III). Pharmacoepidemiol Drug Saf. 2003;12(4):315–26.
ACE: Inhibitors angiotensin-converting-enzyme inhibitor; ADEs: Adverse drug 2. Non-steroidal anti-inflammatory drugs and serious gastrointestinal adverse
events; AIN: Interstitial nephritis; AKI: Acute kidney injury; ASA: Acetylsalicylic reactions–2. British Med J (Clinical research ed) 1986, 292(6529):1190.
acid; CI: Confidence interval; CKD: Chronic kidney disease; CrCl: Creatinine 3. Fries JF, Williams CA, Bloch DA, Michel BA. Nonsteroidal anti-inflammatory
clearance; GFR: Glomerular filtration rate; ICD: International Classification of drug-associated gastropathy: incidence and risk factor models. Am J Med.
Diseases; KDIGO: Kidney Disease Improving Global Outcomes; 1991;91(3):213–22.
MOOSE: Meta-analysis of observational studies in epidemiology; 4. Howard RL, Avery AJ, Slavenburg S, Royal S, Pipe G, Lucassen P,
NSAIDs: Non-steroidal anti-inflammatory drugs; OR: Odds ratio; OTC: Over Pirmohamed M. Which drugs cause preventable admissions to hospital? A
the counter; RR: Relative risk; SCr: Serum creatinine; UK: United Kingdom; systematic review. Br J Clin Pharmacol. 2007;63(2):136–47.
US: United States 5. Coxib and traditional NSAID Trialists' (CNT) Collaboration. Vascular and
upper gastrointestinal effects of non-steroidal anti-inflammatory drugs:
Acknowledgments meta-analyses of individual participant data from randomised trials. Lancet.
The authors thank Mr. Andy Jackson and Ms. Denise Melville (librarians) from 2013;382(9894):769–79.
University of Dundee for training assistance for searching to XZ. 6. Leonard CE, Freeman CP, Newcomb CW, Reese PP, Herlim M, Bilker WB,
Part of this work has been presented in the form of a poster at the 42nd Hennessy S, Strom BL. Proton pump inhibitors and traditional nonsteroidal
North American Primary Care Research Group Annual Meeting held in New anti-inflammatory drugs and the risk of acute interstitial nephritis and acute
York City, New York, November 21 to 25, 2014. kidney injury. Pharmacoepidemiol Drug Safety. 2012;21(11):1155–72.
7. Atkinson M, Basch C, Brett L. Long-term renal and hepatic tolerability of
naproxen: a review of effects in young and elderly patients. Clin Ther.
Funding
1991;13(SUPPL. A):44–50.
This work was supported by a joint PhD scholarship awarded to XZ from
8. Bouvy ML, Heerdink ER, Hoes AW, Leufkens HG. Effects of NSAIDs on the
University of Dundee, School of Medicine and China Scholarship Council.
incidence of hospitalisations for renal dysfunction in users of ACE inhibitors.
The funders had no role in any stage of the design and conduct of the
Drug Saf. 2003;26(13):983–9.
study, collection, management, analysis, and interpretation of data in the
9. Murray MD, Brater DC, Tierney WM, Hui SL, McDonald CJ. Ibuprofen-
study, or the preparation, review, or approval of the manuscript.
associated renal impairment in a large general internal medicine practice.
Am J Med Sci. 1990;299(4):222–9.
Availability of data and materials 10. Patel K, Diamantidis C, Zhan M, Hsu VD, Walker LD, Gardner J, Weir MR, Fink JC.
All data generated or analysed during this study are included in this published
Influence of creatinine versus glomerular filtration rate on non-steroidal
article [and its Additional files].
anti-inflammatory drug prescriptions in chronic kidney disease. Am J Nephrol.
2012;36(1):19–26.
Authors’ contributions 11. Hull S, Mathur R, Dreyer G, Yaqoob MM. Evaluating ethnic differences in the
XZ screened all of the citations for eligibility based on title and abstract, prescription of NSAIDs for chronic kidney disease: a cross-sectional survey of
performed study selection and quality assessment from full-text papers patients in general practice. Br J Gen Pract. 2014;64(624):e448–55.
retrieved, conducted the meta-analyses, and was a major contributor in 12. Adams RJ, Appleton SL, Gill TK, Taylor AW, Wilson DH, Hill CL. Cause for
writing the manuscript. SB screened part of the citations for eligibility concern in the use of non-steroidal anti-inflammatory medications in the
based on title and abstract. BG screened part of the citations for eligibility community–a population-based study. BMC Fam Pract. 2011;12:70.
based on title and abstract and performed study selection and quality 13. Silverstein FE, Faich G, Goldstein JL, Simon LS, Pincus T, Whelton A, Makuch R,
assessment from full-text papers retrieved. PTD provided with statistical Eisen G, Agrawal NM, Stenson WF, et al. Gastrointestinal toxicity with celecoxib
advice on the meta-analyses and resolved disagreements between XZ vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid
and BG when assessing quality for full-text papers as a third reviewer. All arthritis: the CLASS study: a randomized controlled trial. Celecoxib Long-term
authors read and approved the final manuscript. Arthritis Safety Study. JAMA. 2000;284(10):1247–55.
14. Catella-Lawson F, McAdam B, Morrison BW, Kapoor S, Kujubu D, Antes L,
Ethics approval and consent to participate Lasseter KC, Quan H, Gertz BJ, FitzGerald GA. Effects of specific inhibition of
Not applicable. cyclooxygenase-2 on sodium balance, hemodynamics, and vasoactive
eicosanoids. J Pharmacol Exp Ther. 1999;289(2):735–41.
Consent for publication 15. Whelton A, Schulman G, Wallemark C, Drower EJ, Isakson PC, Verburg KM,
Not applicable. Geis GS. Effects of celecoxib and naproxen on renal function in the elderly.
Arch Intern Med. 2000;160(10):1465–70.
Competing interests 16. Swan SK, Rudy DW, Lasseter KC, Ryan CF, Buechel KL, Lambrecht LJ, Pinto MB,
PTD is a member of the New Drugs Committee of the Scottish Medicines Dilzer SC, Obrda O, Sundblad KJ, et al. Effect of cyclooxygenase-2 inhibition on
Consortium (the government body recommending whether drugs be used renal function in elderly persons receiving a low-salt diet. A randomized,
in NHS Scotland). PTD has received grants from GSK, Novo Nordisk, Gilead controlled trial. Ann Intern Med. 2000;133(1):1–9.
Zhang et al. BMC Nephrology (2017) 18:256 Page 12 of 12

17. Whelton A, Maurath CJ, Verburg KM, Geis GS. Renal safety and tolerability of 42. Nderitu P, Doos L, Jones PW, Davies SJ, Kadam UT. Non-steroidal anti-
celecoxib, a novel cyclooxygenase-2 inhibitor. Am J Ther. 2000;7(3):159–75. inflammatory drugs and chronic kidney disease progression: a systematic
18. Marks JL, Colebatch AN, Buchbinder R, Edwards CJ. Pain management for review. Fam Pract. 2013;30(3):247–55.
rheumatoid arthritis and cardiovascular or renal comorbidity. Cochrane 43. Group KDIGOKAKIW. KDIGO clinical practice guideline for acute kidney
Database Systematic Reviews. 2011;10:CD008952. injury. Kidney Inter, Suppl. 2012;2:1–138.
19. Lee A, Cooper MG, Craig JC, Knight JF, Keneally JP. Effects of nonsteroidal 44. Kidney Disease: Improving Global Outcomes. KDIGO Clinical Practice
anti-inflammatory drugs on postoperative renal function in adults with normal Guideline for Acute Kidney Injury Section 2: AKI Definition. Kidney
renal function. Cochrane Database Systematic Reviews. 2007;2:CD002765. International Supplements. 2012;2(Supplement 1):124-128.
20. British Medical Association, Royal Pharmaceutical Society. British national 45. Dreischulte T, Donnan P, Grant A, Hapca A, McCowan C, Guthrie B. Safer
formulary. 63rd ed. London: BMJ Group, Pharmaceutical Press; 2012. prescribing — a trial of education, informatics, and financial incentives. N
21. The Newcastle-Ottawa Scale (NOS) for assessing the quality if nonrandomized Engl J Med. 2016;374(11):1053–64.
studies in meta-analyses [http://www.ohri.ca/programs/clinical_epidemiology/ 46. Guthrie B, Kavanagh K, Robertson C, Barnett K, Treweek S, Petrie D, Ritchie L,
oxford.htm]. Bennie M. Data feedback and behavioural change intervention to improve
22. Deeks JJ, Higgins JPT, Altman DG. Analysing Data and Undertaking Meta- primary care prescribing safety (EFIPPS): multicentre, three arm, cluster
Analyses. In: Cochrane Handbook for Systematic Reviews of Interventions. randomised controlled trial. BMJ. 2016;354:i4079.
edn. Chichester: John Wiley & Sons, Ltd; 2008: 243–296. 47. Dreischulte T, Morales DR, Bell S, Guthrie B. Combined use of nonsteroidal
23. Reeves BC, Deeks JJ, Higgins JPT, Wells GA. Including Non-Randomized anti-inflammatory drugs with diuretics and/or renin-angiotensin system
Studies. In: Cochrane Handbook for Systematic Reviews of Interventions. inhibitors in the community increases the risk of acute kidney injury. Kidney
edn. Chichester: John Wiley & Sons, Ltd; 2008: 389–432. Int. 2015;88:396–03.
24. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency
in meta-analyses. BMJ. 2003;327(7414):557–60.
25. Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected
by a simple, graphical test, vol. 315; 1997.
26. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies
in epidemiology: a proposal for reporting. JAMA. 2000;283(15):2008–12.
27. Adam WR. Non-steroidal anti-inflammatory drugs and the risks of acute
renal failure: number needed to harm. Nephrology. 2011;16(2):154–5.
28. Beard K, Lawson DH, MacFarlane GJ. Non-steroidal anti-inflammatory drugs
and acute renal disease: a case control study. Pharmacoepidemiol Drug Saf.
1992;1(1):3–9.
29. Hemmelgarn BR, Culleton BF, Ghali WA. Derivation and validation of a clinical
index for prediction of rapid progression of kidney dysfunction, vol. 100; 2007.
30. Guess H, West R, Strand L, Helston D, Lydick E, Bergman U, Wolski K:
Hospitalizations for renal impairment among users and non-users of non-
steroidal anti-inflammatory drugs in Saskatchewan, Canada, 1983. In: Side-
effects of anti-inflammatory drugs. edn.: Dordrecht:Springer; 1987: 367–375.
31. Evans JMM, McGregor E, McMahon AD, McGilchrist MM, Jones MC, White G,
McDevitt DG, MacDonald TM. Non-steroidal anti-inflammatory drugs and
hospitalization for acute renal failure. QJM. 1995;88(8):551–7.
32. Griffin MR, Yared A, Ray WA. Nonsteroidal antiinflammatory drugs and acute
renal failure in elderly persons. Am J Epidemiol. 2000;151(5):488–96.
33. Henry D, Page J, Whyte I, Nanra R, Hall C. Consumption of non-steroidal
anti-inflammatory drugs and the development of functional renal
impairment in elderly subjects. Results of a case-control study. Br J Clin
Pharmacol. 1997;44(1):85–90.
34. 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(3):531–9.
35. Lafrance JP, Miller DR. Selective and non-selective non-steroidal anti-
inflammatory drugs and the risk of acute kidney injury. Pharmacoepidemiol
Drug Safety. 2009;18(10):923–31.
36. Perez Gutthann S, Garcia Rodriguez LA, Raiford DS, Duque Oliart A, Ris
Romeu J. Nonsteroidal anti-inflammatory drugs and the risk of
hospitalization for acute renal failure. Arch Intern Med. 1996;156(21):2433–9.
37. Schneider V, Levesque LE, Zhang B, Hutchinson T, Brophy JM. Association of
selective and conventional nonsteroidal antiinflammatory drugs with acute
renal failure: a population-based, nested case-control analysis. Am J
Epidemiol. 2006;164(9):881–9. Submit your next manuscript to BioMed Central
38. Warner TD, Giuliano F, Vojnovic I, Bukasa A, Mitchell JA, Vane JR. Nonsteroid and we will help you at every step:
drug selectivities for cyclo-oxygenase-1 rather than cyclo-oxygenase-2 are
associated with human gastrointestinal toxicity: a full in vitro analysis. Proc • We accept pre-submission inquiries
Natl Acad Sci. 1999;96(13):7563–8. • Our selector tool helps you to find the most relevant journal
39. Dieppe P, Bartlett C, Davey P, Doyal L, Ebrahim S. Balancing benefits and
• We provide round the clock customer support
harms: the example of non-steroidal anti-inflammatory drugs, vol. 329; 2004.
40. Ungprasert P, Cheungpasitporn W, Crowson CS, Matteson EL. Individual • Convenient online submission
non-steroidal anti-inflammatory drugs and risk of acute kidney injury: a • Thorough peer review
systematic review and meta-analysis of observational studies. Eur J Internal
• Inclusion in PubMed and all major indexing services
Med. 2015;26(4):285–91.
41. Asghar W, Jamali F. The effect of COX-2-selective meloxicam on the • Maximum visibility for your research
myocardial, vascular and renal risks: a systematic review. Inflammopharmacology.
2015;23(1):1–16. Submit your manuscript at
www.biomedcentral.com/submit