Effect of Aspirin on Vascular and Nonvascular Outcomes
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
REVIEW ARTICLE
ONLINE FIRST
Effect of Aspirin on Vascular
and Nonvascular Outcomes
Meta-analysis of Randomized Controlled Trials
Sreenivasa Rao Kondapally Seshasai, MD, MPhil; Shanelle Wijesuriya, MA, MBBChir;
Rupa Sivakumaran, MA, MBBChir; Sarah Nethercott, MA, MBBChir;
Sebhat Erqou, MD, PhD; Naveed Sattar, MD, PhD; Kausik K. Ray, MD
Background: The net benefit of aspirin in prevention of 0.96; number needed to treat, 120), driven primarily by
CVD and nonvascular events remains unclear. Our objec- reduction in nonfatal MI (OR, 0.80; 95% CI, 0.67-0.96; num-
tive was to assess the impact (and safety) of aspirin on vas- ber needed to treat, 162). There was no significant reduc-
cular and nonvascular outcomes in primary prevention. tion in CVD death (OR, 0.99; 95% CI, 0.85-1.15) or can-
cer mortality (OR, 0.93; 95% CI, 0.84-1.03), and there was
Data Sources: MEDLINE, Cochrane Library of Clini- increased risk of nontrivial bleeding events (OR, 1.31; 95%
cal Trials (up to June 2011) and unpublished trial data CI, 1.14-1.50; number needed to harm, 73). Significant
from investigators. heterogeneity was observed for coronary heart disease and
bleeding outcomes, which could not be accounted for by
Study Selection: Nine randomized placebo- major demographic or participant characteristics.
controlled trials with at least 1000 participants each, re-
porting on cardiovascular disease (CVD), nonvascular out- Conclusions: Despite important reductions in nonfatal
comes, or death were included. MI, aspirin prophylaxis in people without prior CVD does
not lead to reductions in either cardiovascular death or can-
Data Extraction: Three authors abstracted data. Study- cer mortality. Because the benefits are further offset by clini-
specific odds ratios (ORs) were combined using random- cally important bleeding events, routine use of aspirin for
effectsmeta-analysis.Risksvsbenefitswereevaluatedbycom- primary prevention is not warranted and treatment deci-
paring CVD risk reductions with increases in bleeding. sions need to be considered on a case-by-case basis.
Results: During a mean (SD) follow-up of 6.0 (2.1) years Arch Intern Med. 2012;172(3):209-216.
involving over 100 000 participants, aspirin treatment re- Published online January 9, 2012.
duced total CVD events by 10% (OR, 0.90; 95% CI, 0.85- doi:10.1001/archinternmed.2011.628
W
HILE META - ANALY - guidelines for use of aspirin in primary
Author Affiliations: Cardiac
and Vascular Sciences Research ses to date1,2 have prevention of CVD are based on informa-
Centre, St George’s University shown modest ben- tion from trials published up to 2005,5,6
of London, London, England efits of aspirin for since when at least 3 additional studies
(Drs Seshasai and Ray); the primary preven-
Department of Geriatrics, tion of cardiovascular disease (CVD), it re-
Addenbrooke’s Hospital, mains unclear to what extent these ben- See Invited Commentary
Cambridge, England efits are offset by clinically important at end of article
(Dr Wijesuriya); Department of
Paediatrics, Chelsea and
Westminster Hospital, London CME available online at have been reported.7-9 In this meta-analy-
(Dr Sivakumaran); Department www.jamaarchivescme.com sis we therefore aimed to provide an up-
of Paediatrics, Broomfield and questions on page 208 dated synthesis of evidence regarding the
Hospital, Chelmsford, England wider role of aspirin in primary preven-
(Dr Nethercott); Department of bleeding episodes. Emerging data from pri- tion, including its effect on hitherto un-
Internal Medicine, University of mary and secondary prevention trials also derinvestigated outcomes such as non-
Pittsburgh Medical Center, suggest significant reductions in cancer
Pittsburgh, Pennsylvania vascular disorders (especially cancer),
(Dr Erqou); and Glasgow
mortality in people receiving aspirin pro- and to assess whether the risks vs ben-
Cardiovascular Research phylaxis,3 stimulating discussions for efits of aspirin treatment vary impor-
Centre, University of Glasgow, more widespread use of this agent tantly according to key demographic or
Glasgow, Scotland (Dr Sattar). among healthy individuals.4 Current participant characteristics.10-12
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
209
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012ally unavailable in primary trial reports,
we obtained relevant information from
Detailed search of Cochrane and MEDLINE using terms Aspirin AND
Primary Prevention AND Cardiovascular Disease AND Mortality OR
(1) subsequent trial reports that had pub-
Aspirin AND Primary Prevention AND Cancer lishedinformationonnonvascularevents13;
(2) a recent individual-participant data
meta-analysis of aspirin in mixed popu-
680 Potentially relevant articles (610 from MEDLINE and 70 from lations3 (ie, including both primary and
the Cochrane Library) secondary prevention populations) and
using numbers provided therein to derive
445 Articles were excluded because they were reviews data on additional end points like noncan-
or not clinical trials cer, nonvascular death; (3) investigators
of individual studies (2 studies [Hyperten-
14 Citations were excluded because they overlapped sionOptimalTreatmentTrial{HOT}14 and
between databases Physicians’HealthStudy{PHS}15]provided
previously unpublished data on cancer).
44 Citations tested interventions other than aspirin As nonvascular outcomes were generally
and therefore were excluded reportedasfatalevents,riskestimateswere
calculated for cancer and other nonvas-
11 Citations were excluded because they were not cularmortalityratherthanincidence.Trials
placebo-controlled trials that enrolled subjects with pre-existing
PAD were eligible for inclusion if they had
8 Citations were excluded because they ascertained been asymptomatic for this condition and
outcomes not relevant to this meta-analysis had no history of CVD. Trials of second-
ary prevention or mixed primary and sec-
77 Citations were excluded because they were not ondary prevention,16 pilot studies,17 and
primary prevention trials studies comparing aspirin with other an-
tiplatelet agents instead of placebo18 were
17 Studies were excluded because they were not excluded. In case of multiple publications
randomized controlled trials from the same source, we used informa-
tion from the primary trial report unless
2 Studies were identified and included through hand- stated otherwise. Thus, 9 trials involving
searching reference lists of other citations 102 621 participants were eligible for the
meta-analysis.
66 Citations were related to relevant primary prevention Three authors (S.W., R.S., and S.N.)
trials testing aspirin and cardiovascular outcomes or independentlyabstractedthedata(includ-
cancer outcomes ing demographic characteristics, number
of participants and events, mean [or me-
54 Citations were follow-up publications related to dian] follow-up duration, and risk esti-
original trials and therefore were excluded mates), and discrepancies were resolved
through discussion (S.R.K.S. and K.K.R.).
12 Original publications were related to aspirin in For studies that reported combined clini-
primary prevention of cardiovascular disease cal end points and at least 1 subsidiary
end point (eg, total CHD and either non-
ASPREE was ongoing study and was therefore excluded; fatal MI or fatal CHD but not both), the
PACE and LASAF were excluded because they did not numberofeventsforthemissingendpoint
fulfill inclusion criteria
were calculated by simple subtraction (or
addition, as relevant), assuming that these
9 Original trials were included in our meta-analysis: events did not overlap. Our primary ef-
PHS,23 BDS,22 HOT,24 PPP,26 TPT,25 WHS,13 ficacy end points were total CHD and total
POPADAD,7 JPAD,8 and AAA9
cancer mortality, with the secondary ef-
ficacy end points being subtypes of vas-
Figure 1. Details of literature review. See Abbreviations footnote in the Table for a list of the trial names. cular disease, total CVD events, cause-
ASPREE indicates Aspirin in Reducing Events in the Elderly; LASAF, Low-Dose Aspirin, Stroke, Atrial specific death, and all-cause mortality.
Fibrillation; and PACE, Prevention With Low-Dose Aspirin of Cardiovascular Disease in the Elderly. Because definitions for major bleeding
events varied across studies, and since
participant-level data were unavailable to
METHODS inclusion criteria were randomized
allow reclassification according to stan-
placebo-controlledtrialsthathadincluded dard criteria,19,20 we defined a category of
We searched the electronic databases at least 1000 participants (without previ- clinically “nontrivial” bleeding (fatal
PubMed and Cochrane Library from their ous CHD or stroke, ie, primary prevention bleeding from any site; cerebrovascular
inception to June 2011 using terms re- studies) and had at least 1 year of follow- or retinal bleeding; bleeding from hollow
lated to aspirin, coronary heart disease up during which CHD and/or CVD out- viscus; bleeding requiring hospitalization
(CHD), CVD, cancer, nonvascular events, comes (CHD, stroke, cerebrovascular dis- and/or transfusion; or study-defined ma-
all-cause mortality, clinical trials, and pri- ease, heart failure, and peripheral arterial jor bleeding regardless of source) as our
mary prevention, without restriction to disease [PAD]) were recorded as the main composite primary safety end point. This
anylanguage(Figure 1).Thiswassupple- end points, and details were provided of roughly corresponds to type 2 or above
mented by hand-searching their reference bleeding events. As data on cancer and of the Bleeding Academic Research Con-
lists for additional studies. Our predefined other nonvascular outcomes were gener- sortium definition for bleeding.21
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
210
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012To assess the effect of aspirin we cal- cal analyses were performed using Stata events; 3.9 vs 4.0 for CVD mortal-
culated study-specific unadjusted odds (version 10.1) software (StataCorp). ity; 6.6 vs 7.2 for non-CVD death;
ratios (ORs) before combining them using 5.3 vs 5.9 for cancer death; 3.1 vs 3.2
random-effectsmeta-analysis(fixed-effect for noncancer, nonvascular death;
meta-analyses were conducted for com- RESULTS
11.0 vs 11.7 for all-cause mortality;
parison). We used calculated ORs instead
of reported hazard ratios (HRs) to maxi-
36.0 vs 21.2 for total bleeding events;
STUDY POPULATION and 9.7 vs 7.4 for nontrivial bleed-
mize available data on individual end
points, and for consistency. Given the rare ing events. Losses to follow-up in in-
occurrence of many outcomes in primary
Nine good-quality randomized con- dividual studies are summarized in
prevention, we assumed that the calcu- trolled trials of aspirin for primary eTable 2.
lated ORs would closely approximate re- prevention of CVD including 102 621
ported HRs. Because individual studies participants were eligible (Figure 1;
EFFECTS OF ASPIRIN
differed with regard to various character- T a b l e ; and eTable 1 [http://
ON VASCULAR AND
istics, heterogeneity was quantified using www.archinternmed.com]).7-9,14,15,24-27
NONVASCULAR OUTCOMES
the I2 statistic,22 and potential sources of Most studies were conducted in
heterogeneity were explored by subgroup Western populations and tended to Aspirin treatment was associated with
analyses and metaregression. The I2 sta- include occupational groups (mainly
tistic measures the proportion of overall
a significant 10% reduction in risk of
health professionals15,24,27). Pooled total CVD events (OR, 0.90; 95% CI,
variation in effect estimates that is attrib- weighted mean (SD) age at baseline
utable to between-study heterogeneity. 0.85-0.96), largely owing to a 20% re-
Subgroup analyses involved grouping
for all participants combined was 57 duction in risk of nonfatal MI (OR,
studies according to predefined charac- (4) years, and 46% (n=47 070) were 0.80; 95% CI, 0.67-0.96) (Figure 2).
teristics and calculating stratum-specific male. Although most trials selected There was no beneficial effect on fa-
ORs using random-effects meta-analysis. participants at increased risk for CVD, tal MI (OR, 1.06; 95% CI, 0.83-1.37),
As analyses involved aggregate (and not they did not generally preselect in- stroke (OR, 0.94; 95% CI, 0.84-1.06),
individual-participant) data, it was not dividuals on the basis of diabetes (ex- orCVDdeath(OR,0.99;95%CI,0.85-
possible to study effect-modification by cept Prevention of Progression of Ar- 1.15). Modest, but nonsignificant, re-
various participant-level characteristics. terial Disease And Diabetes Study
Metaregression was used instead to ex-
ductions were observed for total CHD
[POPADAD]7 and Japanese Primary (OR, 0.86; 95% CI, 0.74-1.01), total
plore heterogeneity, using trial-level in- Prevention of Atherosclerosis With
formation. Crude event rates for aspirin nonvascular mortality (OR, 0.92; 95%
and control groups were calculated using
Aspirin for Diabetes Trial [JPAD]8). CI, 0.85-1.00), and all-cause mortal-
data on number of events and mean Other characteristics, including risk ity (OR, 0.94; 95% CI, 0.88-1.00), al-
follow-uptime(whenmeanfollow-upwas factors for CVD, varied widely across though there was no convincing evi-
unavailable, median duration was used studies. dence of benefit with regard to cancer
instead). To contextualize net benefit due mortality (OR, 0.93; 95% CI, 0.84-
to aspirin treatment, we compared rates FOLLOW-UP AND EVENTS 1.03). By contrast, there was a 70%
ofanystatisticallymeaningfulassociations excess risk of total bleeding events
(CVD or nonfatal MI) with rates of bleed-
ing. Mean baseline event rates for the com-
Over a mean (SD) follow-up of 6.0 (OR, 1.70; 95% CI, 1.17-2.46) and a
bined study population were estimated (2.1) years (approximately 700 000 higher than 30% excess risk of non-
by pooling study-specific control event person-years at-risk) in 9 studies, trivial bleeding events (OR, 1.31; 95%
rates for each outcome using random- 2169 CHD events were accrued, of CI, 1.14-1.50) in people receiving as-
effects meta-analysis. Numbers needed which 1540 were nonfatal MI and pirin (eTable 3 contains details of defi-
to treat (NNT) and harm (NNH) were de- 592 were fatal CHD events. One nitions for bleeding). Qualitatively
rived by applying pooled ORs to the mean study8 did not register any fatal MI similar findings were observed in
baseline event rates for the combined events in the aspirin group; hence, analyses restricted to studies of daily
study population. Values of NNT and 0.5 events were added to both treat- aspirin use (ie, after excluding Wom-
NNH provided herein represent the num- ment groups to calculate ORs. Other en’s Health Study [WHS] 13 and
ber of persons that need to be treated with
aspirin for 6 years (the overall mean
major outcomes included stroke PHS15), except that the risk of non-
follow-up time in this study) to avert or (n = 1504); total CVD events trivial bleeding was even higher in
incur, respectively, 1 event. Quality of (n = 4278); CVD death (n = 1285); these studies (OR, 1.48; 95% CI, 1.17-
studies was evaluated using a Delphi scor- nonvascular death (n=2587); can- 1.86; eFigure 1). Considerable hetero-
ing system,23 which is based on the fol- cer death (n=1512, 8 studies); non- geneity was observed for the ORs for
lowing: adequacy of randomization; al- cancer, nonvascular death (n=983, major efficacy and safety end points
location concealment; balance between 8 studies); all-cause mortality (Figure 2 and eFigure 2), which could
randomized groups at baseline; a priori (n = 3895); total bleeding events not be explained by reported charac-
identification of inclusion criteria; pres- (n = 40 712); and nontrivial bleed- teristics (Figure 3 and eFigures 3-5).
ence or absence of blinding; use of ing events (n=10 049). Pooled event The risk of CVD events in people
intention-to-treat analyses; and report-
ing of point estimates and measures of
rates per 1000 person-years of fol- treated with aspirin was, however,
variability for main outcomes. Potential low-up in people randomized to as- lower at an older age (eFigure 6), and
publication bias was investigated using pirin vs placebo were 4.1 vs 5.1 for that of nontrivial bleeding was some-
funnel plots and the Egger test. All P val- nonfatal MI; 1.9 vs 1.9 for fatal MI; what higher at a younger age and at
ues reported are 2-sided; P⬍.05 was con- 7.0 vs 8.1 for total CHD; 3.8 vs 4.0 higher systolic blood pressure (eFig-
sidered statistically significant. Statisti- for stroke; 12.8 vs 14.1 for total CVD ure 5). Contrary to previous re-
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
211
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012Table. Characteristics of Individual Trials Contributing to the Current Analysis
SBP,
No. of Age, Mean Mean (SD),
Source Location Year Participants (SD), y Male, % Diabetes, % Smokers, % mm Hg
BDS 24 England 1988 5139 63.6 100 2 31 135.8
PHS 15 US 1989 22 071 53.8 100 2 11 128.5
HOT 14 Multiple 1998 18 790 61.5 53 8 16 170
TPT 25 UK 1998 5085 57.5 100 NS 41 139
PPP 26 Italy 2001 4495 64.4 42 17 15 145.1
WHS 13 US 2005 39 876 54.6 0 3 13 127.3
POPADAD 7 Scotland 2008 1276 60.3 44 100 31 145
JPAD 8 Japan 2008 2539 64.5 55 100 21 135
AAA 9 Scotland 2010 3350 61.6 28 3 32 147.5
Total or 102 621 57.3 a (4.1) 46 8 16 138 a (17)
Mean (SD)
Total All Aspirin Arm, Placebo Arm,
Cholesterol, Aspirin Dose, Participants, Duration of Duration of
Mean (SD), mg, and Aspirin Concomitant Duration of Follow-up, Follow-up,
Source mmol/L b Schedule Formulation Treatment b Follow-up, y c Person-years d Person-years d
BDS 24 NS 500 or 300 daily Ordinary, soluble or No 6.0 18 820 9470
effervescent (500 mg) or
enteric coated (300 mg)
PHS 15 5.46 325 alternate day Regular (most) No 5.02 54 560 54 356
HOT 14 6.1 75 daily NS Yes 3.8 35 716 35 686
TPT 25 6.4 75 daily Controlled release Yes 6.4 8105 8071
PPP 26 6.1 100 daily Enteric coated Yes 3.6 8014 8168
WHS 13 5.2 100 alternate day NS Yes 10.1 201 333 201 414
POPADAD 7 5.52 100 daily NS Yes 6.7 4275 4275
JPAD 8 5.21 81 or 100 daily NS No 4.37 5515 5580
AAA 9 6.2 a 100 daily Enteric coated No 8.2 13 735 13 735
Total or Mean (SD) 5.5 (0.5) 6.0 (2.1) 350 073 340 755
Abbreviations: AAA, Aspirin for Asymptomatic Atherosclerosis Trial 9; BDS, British Doctors Study 24; HOT, Hypertension Optimal Treatment Trial 14; JPAD, Japanese
Primary Prevention of Atherosclerosis With Aspirin for Diabetes Trial 8; NS, not stated; PHS, Physicians’ Health Study 15; POPADAD, Prevention of Progression of Arterial
Disease and Diabetes Trial 7; PPP, Primary Prevention Project 26; SBP, systolic blood pressure; TPT, Thrombosis Prevention Trial 25; WHS, Women’s Health Study. 13
Conventional unit conversion factor: To convert cholesterol to milligrams per deciliter, divide by 0.0259.
a Represents weighted mean (SD).
b Concomitant treatments include agents other than anti-platelet drugs (eg, blood pressure-lowering medication), as in factorial trials.
c Follow-up duration shown for POPADAD and JPAD represents median follow-up, not mean. Also, total cholesterol values for POPADAD are median, not mean. Data
on cholesterol measurements at baseline were missing in approximately 0.6% of all participants in the AAA study.
d Follow-up duration shown in person-years according to treatment arm was obtained directly from study reports for BDS and TPT, and was calculated based on
numbers per group multiplied by mean (or median) follow-up time for other studies. In PHS, the reported duration of follow-up differed for various outcomes, and the
numbers shown correspond to those for MI (including nonfatal and fatal MI).
ports,28 we did not find any signifi- Aspirin for Asymptomatic Athero- events increased proportionately with
cant sex differences in treatment effect sclerosis Trial [AAA]9), or health care background event rates for these out-
for total CVD events (eFigure 7). professionals (British Doctors Study comes, although the benefit ap-
Lastly, no significant heterogeneity [BDS],8 PHS,15 and WHS13) were ex- peared to be more modest for CVD
between studies was observed for cluded (eTable 4). Results were also than nonfatal MI (Figure 4). Such
nonvascular, cancer, and all-cause similar when fixed-effect meta- benefits were offset by increased rates
mortality (P⬎.10; Figure 2). analysis was used instead of random- of nontrivial bleeding, even though for
effects models. There was no evi- nonfatal MI there was a suggestion
SENSITIVITY ANALYSES dence of publication bias (Egger test that at high baseline event rates there
The effect of aspirin on nonfatal MI P value ⬎.05 for all major out- may be net benefit in favor of aspirin
or total CVD events was unrelated to comes; eFigure 8). prophylaxis. The NNT to avoid 1
its average daily dose and was more nonfatal MI event over 6 years was
pronounced in trials published be- COMPARATIVE MERITS 162 (NNT was 120 to avert 1 CVD
fore 2000 (compared with more re- OF ASPIRIN event over the same period). By com-
cent studies; Figure 3). Findings were parison, the NNT for nonvascular
comparable when studies con- The net benefit due to aspirin treat- death was 292 (247 for cancer death),
ducted exclusively in non-Western ment (expressed as a difference be- and at least 1 nontrivial bleeding event
populations (JPAD8), or people with tween absolute event rates in the pla- was caused for every 73 persons
diabetes (JPAD8 and POPADAD7) or cebo and aspirin treatment arms) for treated with aspirin for approxi-
asymptomatic PAD (POPADAD7 and both nonfatal MI and total CVD mately 6 years.
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
212
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012No. of No. of cases/ No. of cases/
Outcome studies participants, aspirin participants, placebo Odds ratio (95% CI) I2 (95% CI)
Nonfatal MI 9 699/52 145 841/50 476 0.80 (0.67-0.96) 62.1% (21.7%-81.6%)
Fatal MI 9 329/52 145 263/50 476 1.06 (0.83-1.37) 37.4% (0.0%-71.2%)
Total CHD 9 1044/52 145 1125/50 476 0.86 (0.74-1.01) 64.8% (28.1%-82.7%)
Stroke 9 749/52 145 755/50 476 0.94 (0.84-1.06) 14.8% (0.0%-56.9%)
Total CVD events 9 2107/52 145 2171/50 476 0.90 (0.85-0.96) 00.0% (0.0%-55.4%)
CVD mortality 9 674/52 145 611/50 476 0.99 (0.85-1.15) 36.1% (0.0%-70.6%)
Non-CVD mortality 9 1276/52 145 1311/50 476 0.92 (0.85-1.00) 00.0% (0.0%-4.2%)
Cancer mortality 8 750/49 919 762/48 207 0.93 (0.84-1.03) 00.0% (0.0%-49.4%)
Noncancer, nonvascular mortality 8 481/49 919 502/48 207 0.90 (0.76-1.07) 32.1% (0.0%-69.9%)
All-cause mortality 9 1962/52 145 1933/50 476 0.94 (0.88-1.00) 00.0% (0.0%-0.0%)
Total bleeds 9 22 297/50 868 18 415/49 208 1.70 (1.17-2.46) 98.0% (97.3%-98.5%)
Nontrivial bleeds 9 5337/50 868 4712/49 208 1.31 (1.14-1.50) 65.7% (30.3%-83.1%)
0.5 1.0 1.5 2.0 2.5
Odds Ratio
Favors aspirin Favors placebo
Figure 2. Effect of aspirin on vascular and nonvascular outcomes or death. CHD indicates coronary heart disease; CVD, cardiovascular disease; and
MI, myocardial infarction.
Nonfatal MI Total CVD events Nontrivial bleed
Subgroup OR (95% CI) P value OR (95% CI) P value OR (95% CI) P value
Period of publication
After 2000 0.98 (0.84-1.14)regarding cancer mortality. Al-
though recent evidence suggests that
No. of nonfatal MI events averted per 1000 person-years
No. of total CVD events averted per 1000 person-years aspirin reduces mortality from cer-
No. of nontrivial bleeding events caused per 1000 person-years tain cancers,3,29 this is based on in-
Fitted values : nonfatal MI
Fitted values : total CVD events formation from both primary and
A
Fitted values : nontrivial bleeds secondary prevention studies. As
6
TPT background event rates for cancers
and other chronic diseases may be
Absolute Risk Difference per 1000 Person-years
TPT
different for participants with pre-
4 existing CVD, and since concomi-
PHS TPT
tant lifestyle or treatment decisions
PPP
JPAD may alter risks associated with these
JPAD
2 HOT
PHS WHS outcomes, assessments of risk based
PPP PHS HOT on primary prevention trials is likely
HOT
AAA PPP
WHS AAA POPADAD to be more informative. It has also
POPADAD BDS
0
BDS WHS BDS been argued that the frequency of as-
AAA
JPAD
POPADAD
pirin administration is an impor-
tant determinant of cancer out-
–2
comes, with more sustained benefits
0 10 20 30
Absolute Event Rate in Placebo Group per 1000 Person-years
with daily compared with alternate-
B day treatment.3 However, we were
6 TPT unable to confirm these observa-
tions because cancer mortality failed
Absolute No. of Nontrivial Bleeding Events
to reach statistical significance even
Caused per 1000 Person-years
4 after excluding studies that had used
PHS
alternate-day aspirin treatment
(WHS13 and PHS15) (OR, 0.88; 95%
JPAD
2 WHS PPP
HOT CI, 0.76-1.01). It is plausible that
AAA
daily aspirin use may prompt ear-
lier detection of cancers owing to in-
0 BDS
vestigation for bleeding, with appar-
POPADAD ent survival benefits (which may
–2
nevertheless be artifactual).
–2 0 2 4 6
The findings of our analysis merit
Absolute No. of Nonfatal MI Events Averted per 1000 Person-years careful consideration in light of ex-
C isting evidence regarding aspirin use
6 TPT in primary prevention. For in-
stance, the magnitude of risk reduc-
Absolute No. of Nontrivial Bleeding Events
tion observed for nonfatal MI and
Caused per 1000 Person-years
4 total CVD events is broadly consis-
PHS tent with some previous reports.30 It
JPAD
has been suggested that the pharma-
2 WHS HOT
PPP cokinetics of aspirin may be differ-
AAA ent among men and women,31,32 with
0 BDS consequent sex differences in effi-
POPADAD
cacy. However, our analysis did not
reveal any material differences in as-
–2 pirin treatment effect by sex. Al-
0 1 2 3 though these findings may be prone
Absolute No. of Total CVD Events Averted per 1000 Person-years to ecological and other biases, they
are in agreement with large-scale in-
Figure 4. Comparison of risk vs benefit due to aspirin treatment for primary prevention of cardiovascular dividual-participant data meta-
disease. A, Plot of absolute risk difference in relation to background (ie, placebo) event rate for main analyses that showed lack of any im-
outcomes. B, Plot comparing absolute number of nontrivial bleeding events caused vs absolute number portant interaction by sex30 for major
of nonfatal MI events averted. C, Plot comparing absolute number of nontrivial bleeding caused vs
absolute number of total CVD events averted. In each of the panels, data points and associated labels CVD outcomes. Our analysis also
correspond to individual studies, while straight lines represent fitted values. In panel A, the x-axis showed that aspirin was no better
represents the background (ie, placebo) event rate for each of the outcomes of interest (nonfatal MI, total than placebo for reducing nonfatal MI
CVD, and nontrivial bleed), whereas the y-axis shows risk difference for these outcomes (total number of
events averted in case of nonfatal MI and total CVD, as against total number of adverse events caused in
events in trials published after 2000,
case of nontrivial bleeding). In panels B and C, the x-axis shows the absolute number of events averted which may be ascribed to better treat-
for nonfatal MI or total CVD, respectively, in each study plotted against the absolute number of nontrivial ments for CVD or underlying risk fac-
bleed events caused in the same studies. See Abbreviations footnote in the Table for a list of the trial tors. This decline suggests that in
names and reference citations.
contemporary primary prevention
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
214
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012settings, aspirin may add little extra assess the relationship between mer Terrace, London SW17 0RE,
value to other CVD risk reduction aspirin treatment and cancer mor- England (rkondapa@sgul.ac.uk).
strategies that target lipid levels, blood tality. Although this may have some- Author Contributions: Drs Sesha-
pressure, and smoking, especially in what underestimated this associa- sai and Ray had full access to all of
low-risk individuals. tion, it may have in fact been the data in the study and take respon-
On the other hand, aspirin may beneficial for study validity be- sibility for the integrity of the data and
be associated with net harm owing cause estimates based on mortality, the accuracy of the data analysis.
to increased potential for bleeding. rather than incidence, are less likely Study concept and design: Seshasai and
Current guidelines for primary pre- to be affected by ascertainment bias. Ray. Acquisition of data: Seshasai,
vention advocate widespread use of Third, the effect of aspirin on can- Wijesuriya, Sivakumaran, Nether-
aspirin in people at increased risk for cer mortality could be evaluated cott, and Ray. Analysis and interpre-
CVD.5,33,34 Others have even sug- using information from only 8 of 9 tation of data: Seshasai, Wijesuriya,
gested regular prophylaxis in people studies. Nevertheless, these results Sivakumaran, Nethercott, Erqou,
above a certain age, either singly35 are fairly robust because the major- Sattar, and Ray. Drafting of the manu-
or in combination with other ity of primary prevention trials of as- script: Seshasai and Ray. Critical re-
agents.36 However, such strategies re- pirin were included in our analy- vision of the manuscript for impor-
quire closer scrutiny because aspi- ses. Fourth, as we studied the effect tant intellectual content: Seshasai,
rin cannot be compared with either of aspirin on multiple outcomes, Wijesuriya, Sivakumaran, Nether-
statins or blood pressure–lowering some of the associations may be due cott, Erqou, Sattar, and Ray. Statis-
agents with regard to its effects on to chance alone. However, as the risk tical analysis: Seshasai and Erqou.
CVD death. Hence, based on our estimates were largely consistent Study supervision: Seshasai, Sattar,
findings of a marginal benefit on with previous reports,30 the scope of and Ray.
nonfatal MI, a nonsignificant effect any artifactual associations is likely Financial Disclosure: None reported.
on cancer death, and a significantly to be limited. Lastly, as most stud- Additional Contributions: J.
increased risk of clinically relevant ies were performed in occupational Michael Gaziano, MD, MPH, (US
bleeding, it is perhaps timely to re- groups in Western populations, find- Physicians Health Study) and Lar-
appraise existing guidelines for as- ings of this meta-analysis may not rye E. Loss, PharmD, MBA, PMP, of
pirin use in primary prevention. Our be entirely generalizable. AstraZeneca (Hypertension Opti-
data additionally highlight the need In conclusion, we found rather mum Trial) provided tabular data,
for more robust evidence in spe- modest benefits of aspirin treat- and Stephen Kaptoge, PhD, Univer-
cific subgroups of participants,37,38 ment on nonfatal MI and total CVD sity of Cambridge, provided statis-
since current guidelines39 are based events in primary prevention, while tical advice.
on limited evidence in different sub- the effect on cancer mortality was Online-Only Material: The eTables
groups. Future studies should there- nonsignificant. Because the ben- and eFigures are available at http:
fore aim to assess the impact of low- efits of aspirin treatment were ac- //www.archinternmed.com.
dose, alternate-day aspirin treatment companied by a significant in-
on both vascular and nonvascular crease in risk of bleeding, further REFERENCES
outcomes, especially in specific sub- study is needed to identify subsets
groups of individuals40 and within of participant having favorable risk 1. Bartolucci AA, Tendera M, Howard G. Meta-
diverse populations.41 Also, owing to to benefit ratio for aspirin use in pri- analysis of multiple primary prevention trials of
the relatively short mean follow-up mary prevention and/or involving cardiovascular events using aspirin. Am J Cardiol.
2011;107(12):1796-1801.
duration reported in this meta- more high-risk participants. In the 2. Raju N, Sobieraj-Teague M, Hirsh J, O’Donnell
analysis, longer-term studies may be absence of such information, a re- M, Eikelboom J. Effect of aspirin on mortality in
warranted to clarify the precise role appraisal of current guidelines ap- the primary prevention of cardiovascular disease.
of aspirin in cancer prevention. pears to be warranted, particularly Am J Med. 2011;124(7):621-629.
3. Rothwell PM, Fowkes FG, Belch JF, Ogawa H, War-
Despite obvious advantages, there in countries where a large number low CP, Meade TW. Effect of daily aspirin on long-
are important limitations to our of otherwise healthy adults are pre- term risk of death due to cancer: analysis of in-
analyses. First, we were unable to scribed aspirin, since a significant dividual patient data from randomised trials.
harmonize outcome definitions proportion of them may develop Lancet. 2011;377(9759):31-41.
across studies (especially for out- bleeding complications.42 4. Opie LH. Aspirin in the prevention of cancer [letter].
Lancet. 2011;377(9778):1651.
comes with high heterogeneity such 5. US Preventive Services Task Force. Aspirin for the
as bleeding) and were further un- Accepted for Publication: October prevention of cardiovascular disease: US Preven-
able to quantify precisely the effect 25, 2011. tive Services Task Force recommendation statement.
of aspirin treatment in clinically rel- Published Online: January 9, 2012. Ann Intern Med. 2009;150(6):396-404.
6. Wolff T, Miller T, Ko S. Aspirin for the primary pre-
evant subgroups. Nonetheless, we doi:10.1001/archinternmed.2011 vention of cardiovascular events: an update of the
combined bleeding episodes that .628 evidence for the US Preventive Services Task
were unlikely to be trivial and con- Correspondence: Sreenivasa Rao Force. Ann Intern Med. 2009;150(6):405-410.
ducted subgroup analyses using Kondapally Seshasai, MD, MPhil, 7. Belch J, MacCuish A, Campbell I, et al; Preven-
available summary information. Sec- NIHR Clinical Lecturer in Preventa- tion of Progression of Arterial Disease and Dia-
betes Study Group; Diabetes Registry Group; Royal
ond, as data on cancer incidence tive Cardiology, Cardiac and Vascu- College of Physicians Edinburgh. The prevention
were generally unavailable from pub- lar Sciences Research Centre, St of progression of arterial disease and diabetes
lished reports, we were only able to George’s University of London, Cran- (POPADAD) trial: factorial randomised placebo
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
215
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012controlled trial of aspirin and antioxidants in pa- bolysis in Myocardial Infarction (TIMI) Trial, Phase ferences in the effects of aspirin on the interac-
tients with diabetes and asymptomatic periph- I: a comparison between intravenous tissue plas- tion of platelets with subendothelium. Thromb Res.
eral arterial disease. BMJ. October 16, 2008; minogen activator and intravenous streptoki- 1986;44(6):837-847.
337:a1840. nase: clinical findings through hospital discharge. 32. Spranger M, Aspey BS, Harrison MJ. Sex differ-
8. Ogawa H, Nakayama M, Morimoto T, et al; Japa- Circulation. 1987;76(1):142-154. ence in antithrombotic effect of aspirin. Stroke.
nese Primary Prevention of Atherosclerosis With 20. The GUSTO investigators. An international ran- 1989;20(1):34-37.
Aspirin for Diabetes ( JPAD) Trial Investigators. domized trial comparing four thrombolytic strat- 33. British Cardiac Society; British Hypertension
Low-dose aspirin for primary prevention of ath- egies for acute myocardial infarction. N Engl J Med. Society; Diabetes UK; HEART UK; Primary Care
erosclerotic events in patients with type 2 diabe- 1993;329(10):673-682. Cardiovascular Society; Stroke Association.
tes: a randomized controlled trial. JAMA. 2008; 21. Mehran R, Rao SV, Bhatt DL, et al. Standardized JBS 2: Joint British Societies’ guidelines on pre-
300(18):2134-2141. bleeding definitions for cardiovascular clinical vention of cardiovascular disease in clinical
9. Fowkes FG, Price JF, Stewart MC, et al; Aspirin trials: a consensus report from the Bleeding Aca- practice. Heart. 2005;91(suppl 5):v1-v52.
for Asymptomatic Atherosclerosis Trialists. demic Research Consortium. Circulation. 2011; 34. Pearson TA, Blair SN, Daniels SR, et al; American
Aspirin for prevention of cardiovascular events in 123(23):2736-2747. Heart Association Science Advisory and Coordinat-
a general population screened for a low ankle bra- 22. Higgins JP, Thompson SG. Quantifying hetero- ing Committee. AHA Guidelines for Primary Pre-
chial index: a randomized controlled trial. JAMA. geneity in a meta-analysis. Stat Med. 2002;21 vention of Cardiovascular Disease and Stroke: 2002
2010;303(9):841-848. (11):1539-1558. Update: Consensus Panel Guide to Comprehen-
10. Serebruany VL, Steinhubl SR, Berger PB, et al. 23. Verhagen AP, de Vet HC, de Bie RA, et al. The Del- sive Risk Reduction for Adult Patients Without Coro-
Analysis of risk of bleeding complications after dif- phi list: a criteria list for quality assessment of ran-
nary or Other Atherosclerotic Vascular Diseases.
ferent doses of aspirin in 192,036 patients en- domized clinical trials for conducting systematic
Circulation. 2002;106(3):388-391.
rolled in 31 randomized controlled trials. Am J reviews developed by Delphi consensus. J Clin
35. Elwood P, Morgan G, Brown G, Pickering J.
Cardiol. 2005;95(10):1218-1222. Epidemiol. 1998;51(12):1235-1241.
Aspirin for everyone older than 50? BMJ. 2005;
11. Lee M, Cryer B, Feldman M. Dose effects of aspirin 24. Peto R, Gray R, Collins R, et al. Randomised trial of
330(7505):1440-1441.
on gastric prostaglandins and stomach mucosal prophylactic daily aspirin in British male doctors. Br
36. Wald NJ, Law MR. A strategy to reduce cardio-
injury. Ann Intern Med. 1994;120(3):184-189. Med J (Clin Res Ed). 1988;296(6618):313-316.
vascular disease by more than 80%. BMJ. 2003;
12. Pirmohamed M, James S, Meakin S, et al. Ad- 25. The Medical Research Council’s General Prac-
326(7404):1419.
verse drug reactions as cause of admission to hos- tice Research Framework. Thrombosis preven-
37. De Berardis G, Sacco M, Strippoli GF, et al. As-
pital: prospective analysis of 18 820 patients. BMJ. tion trial: randomised trial of low-intensity oral an-
pirin for primary prevention of cardiovascular
2004;329(7456):15-19. ticoagulation with warfarin and low-dose aspirin
13. Cook NR, Lee IM, Gaziano JM, et al. Low-dose as- in the primary prevention of ischaemic heart dis- events in people with diabetes: meta-analysis of
pirin in the primary prevention of cancer: the Wom- ease in men at increased risk. Lancet. 1998; randomised controlled trials. BMJ. November 6,
en’s Health Study: a randomized controlled trial. 351(9098):233-241. 2009;339:b4531.
JAMA. 2005;294(1):47-55. 26. de Gaetano G; Collaborative Group of the Pri- 38. Zhang C, Sun A, Zhang P, et al. Aspirin for pri-
14. Hansson L, Zanchetti A, Carruthers SG, et al; HOT mary Prevention Project. Low-dose aspirin and vi- mary prevention of cardiovascular events in pa-
Study Group. Effects of intensive blood-pressure tamin E in people at cardiovascular risk: a ran- tients with diabetes: A meta-analysis. Diabetes Res
lowering and low-dose aspirin in patients with hy- domised trial in general practice. Lancet. 2001; Clin Pract. 2010;87(2):211-218.
pertension: principal results of the Hypertension Op- 357(9250):89-95. 39. Pignone M, Alberts MJ, Colwell JA, et al. Aspirin
timal Treatment (HOT) randomised trial. Lancet. 27. Ridker PM, Cook NR, Lee IM, et al. A randomized for primary prevention of cardiovascular events
1998;351(9118):1755-1762. trial of low-dose aspirin in the primary preven- in people with diabetes: a position statement of
15. Steering Committee of the Physicians’ Health Study tion of cardiovascular disease in women. N Engl the American Diabetes Association, a scientific
Research Group. Final report on the aspirin com- J Med. 2005;352(13):1293-1304. statement of the American Heart Association, and
ponent of the ongoing Physicians’ Health Study. 28. Berger JS, Roncaglioni MC, Avanzini F, Pan- an expert consensus document of the American
N Engl J Med. 1989;321(3):129-135. grazzi I, Tognoni G, Brown DL. Aspirin for the pri- College of Cardiology Foundation. Circulation.
16. ETDRS Investigators. Aspirin effects on mortal- mary prevention of cardiovascular events in 2010;121(24):2694-2701.
ity and morbidity in patients with diabetes melli- women and men: a sex-specific meta-analysis of 40. De Berardis G, Sacco M, Evangelista V, et al;
tus: Early Treatment Diabetic Retinopathy Study randomized controlled trials. JAMA. 2006;295 ACCEPT-D Study Group. Aspirin and Simvasta-
report 14. JAMA. 1992;268(10):1292-1300. (3):306-313. tin Combination for Cardiovascular Events Pre-
17. Silagy CA, McNeil JJ, Donnan GA, Tonkin AM, 29. Rothwell PM, Wilson M, Elwin CE, et al. Long-term vention Trial in Diabetes (ACCEPT-D): design of
Worsam B, Campion K. The PACE pilot study: 12- effect of aspirin on colorectal cancer incidence and a randomized study of the efficacy of low-dose as-
month results and implications for future pri- mortality: 20-year follow-up of five randomised trials. pirin in the prevention of cardiovascular events in
mary prevention trials in the elderly. (Prevention Lancet. 2010;376(9754):1741-1750. subjects with diabetes mellitus treated with statins.
with low-dose Aspirin of Cardiovascular disease 30. Baigent C, Blackwell L, Collins R, et al; Antithrom- Trials. August 28, 2007;8:21.
in the Elderly). J Am Geriatr Soc. 1994;42(6): botic Trialists’ (ATT) Collaboration. Aspirin in the 41. Aspirin to reduce risk of initial vascular events.
643-647. primary and secondary prevention of vascular dis- ARRIVE Web Site. http://www.arrive-study.com
18. Bhatt DL, Fox KA, Hacke W, et al; CHARISMA In- ease: collaborative meta-analysis of individual par- /EN/. Accessed October 26, 2010.
vestigators. Clopidogrel and aspirin versus aspi- ticipant data from randomised trials. Lancet. 2009; 42. Ajani UA, Ford ES, Greenland KJ, Giles WH, Mok-
rin alone for the prevention of atherothrombotic 373(9678):1849-1860. dad AH. Aspirin use among US adults: behav-
events. N Engl J Med. 2006;354(16):1706-1717. 31. Escolar G, Bastida E, Garrido M, Rodrı́guez- ioral risk factor surveillance system. Am J Prev
19. Chesebro JH, Knatterud G, Roberts R, et al. Throm- Gómez J, Castillo R, Ordinas A. Sex-related dif- Med. 2006;30(1):74-77.
ARCH INTERN MED/ VOL 172 (NO. 3), FEB 13, 2012 WWW.ARCHINTERNMED.COM
216
©2012 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Illinois-Urbana Champaign User on 09/04/2012You can also read
