The n e w e ng l a n d j o u r na l of m e dic i n e

Original Article

Effects of Aspirin for Primary Prevention

in Persons with Diabetes Mellitus
The ASCEND Study Collaborative Group*​​

A BS T R AC T

BACKGROUND
Diabetes mellitus is associated with an increased risk of cardiovascular events. The members of the writing committee
Aspirin use reduces the risk of occlusive vascular events but increases the risk of (Louise Bowman, M.D., Marion Mafham,
M.D., Karl Wallendszus, M.Sc., Will Ste-
bleeding; the balance of benefits and hazards for the prevention of first cardiovas- vens, Ph.D., Georgina Buck, M.Sc., Jill
cular events in patients with diabetes is unclear. Barton, Kevin Murphy, Theingi Aung,
M.D., Richard Haynes, D.M., Jolyon Cox,
METHODS D.Phil., Aleksandra Murawska, M.Sc., Al-
len Young, Ph.D., Michael Lay, D.Phil.,
We randomly assigned adults who had diabetes but no evident cardiovascular dis- Fang Chen, M.D., Ph.D., Emily Sammons,
ease to receive aspirin at a dose of 100 mg daily or matching placebo. The primary M.B., Ch.B., Emma Waters, M.B., B.S.,
efficacy outcome was the first serious vascular event (i.e., myocardial infarction, Amanda Adler, M.D., Ph.D., Jonathan Bo-
dansky, M.D., Andrew Farmer, D.M.,
stroke or transient ischemic attack, or death from any vascular cause, excluding Roger McPherson, B.M., F.R.C.Ophth.,
any confirmed intracranial hemorrhage). The primary safety outcome was the first Andrew Neil, D.Sc., F.R.C.P., David Simp-
major bleeding event (i.e., intracranial hemorrhage, sight-threatening bleeding son, Richard Peto, F.R.S., F.Med.Sci., Co-
lin Baigent, F.F.P.H., F.R.C.P., Rory Collins,
event in the eye, gastrointestinal bleeding, or other serious bleeding). Secondary F.R.S., F.Med.Sci., Sarah Parish, D.Phil.,
outcomes included gastrointestinal tract cancer. and Jane Armitage, F.R.C.P., F.F.P.H.) as-
sume responsibility for the overall con-
RESULTS tent and integrity of this article. The affili-
A total of 15,480 participants underwent randomization. During a mean follow-up ations of the members of the writing
committee are listed in the Appendix.
of 7.4 years, serious vascular events occurred in a significantly lower percentage of Address reprint requests to Dr. Armitage
participants in the aspirin group than in the placebo group (658 participants at the Medical Research Council, Popula-
[8.5%] vs. 743 [9.6%]; rate ratio, 0.88; 95% confidence interval [CI], 0.79 to 0.97; tion Health Research Unit, Clinical Trial
Service Unit and Epidemiological Studies
P = 0.01). In contrast, major bleeding events occurred in 314 participants (4.1%) in Unit, Nuffield Department of Population
the aspirin group, as compared with 245 (3.2%) in the placebo group (rate ratio, Health, Richard Doll Bldg., Old Road
1.29; 95% CI, 1.09 to 1.52; P = 0.003), with most of the excess being gastrointesti- Campus, Roosevelt Dr., Oxford OX3 7LF,
United Kingdom, or at ­ascend@​­ndph​.­ox
nal bleeding and other extracranial bleeding. There was no significant difference ​.­ac​.­uk or ­jane​.­armitage@​­ndph​.­ox​.­ac​.­uk.
between the aspirin group and the placebo group in the incidence of gastrointes-
*A complete list of the members of the
tinal tract cancer (157 participants [2.0%] and 158 [2.0%], respectively) or all cancers ASCEND Study Collaborative Group is
(897 [11.6%] and 887 [11.5%]); long-term follow-up for these outcomes is planned. provided in Supplementary Appendix 1,
available at NEJM.org.
CONCLUSIONS Drs. Bowman and Mafham and Drs. Col-
Aspirin use prevented serious vascular events in persons who had diabetes and no lins, Parish, and Armitage contributed
evident cardiovascular disease at trial entry, but it also caused major bleeding equally to this article.
events. The absolute benefits were largely counterbalanced by the bleeding hazard. This article was published on August 26,
(Funded by the British Heart Foundation and others; ASCEND Current Controlled 2018, at NEJM.org.
Trials number, ISRCTN60635500; ClinicalTrials.gov number, NCT00135226.) DOI: 10.1056/NEJMoa1804988
Copyright © 2018 Massachusetts Medical Society.

n engl j med nejm.org 1
The New England Journal of Medicine
Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

I
t is well established that aspirin use will be useful in assessing any effects of aspirin
is beneficial for patients with cardiovascular use on cancer more reliably. We performed the
disease, but it is less clear that there is over- ASCEND (A Study of Cardiovascular Events in Dia-
all benefit in persons who have not yet had a betes) randomized trial to assess the efficacy and
cardiovascular event.1,2 Patients with diabetes safety of enteric-coated aspirin at a dose of 100 mg
mellitus have a risk of vascular events that is two daily, as compared with placebo, in persons who
to three times as high as the risk among those had diabetes without manifest cardiovascular
without diabetes,3 but most of the estimated 400 disease at trial entry. Using a factorial design in
million persons with diabetes worldwide do not the same trial, we also randomly assigned the pa-
have manifest vascular disease.4 tients to receive a daily regimen of either n−3 fatty
The 2009 Antithrombotic Trialists’ Collabora- acids, administered as 1-g capsules, or placebo,
tion meta-analysis involving 95,000 patients in findings that are now reported elsewhere in the
six primary prevention trials showed that assign- Journal.14
ment to aspirin use led to a 12% (95% confidence
interval [CI], 6 to 18) lower risk of serious vascu- Me thods
lar events than control.1 On average, however, the
approximate 50% higher risk of bleeding with Trial Oversight
aspirin use than with control counterbalanced ASCEND was designed and conducted by inde-
much of the benefit in these low-risk patients. pendent investigators in the Clinical Trial Service
Only approximately 4% of participants in those Unit at the University of Oxford (the regulatory
trials had diabetes, and the lower relative risk trial sponsor). The trial methods, characteristics of
among them was similar to that observed among the participants, and data analysis plan (including
participants without diabetes (as was also ob- outcome definitions) have been reported previ-
served in the context of secondary prevention). ously.15,16 The protocol (available with the full text
Likewise, the higher relative risk of bleeding of this article at NEJM.org) was approved by the
with aspirin use than with control was similar North West Multicenter Research Ethics Com-
among persons with diabetes and those without mittee. The trial was funded by the British Heart
diabetes. Foundation. The aspirin and matching placebo
Since the analyses of the Antithrombotic Tri- (along with funding for packaging) were pro-
alists’ Collaboration, four trials of aspirin use for vided by Bayer (Germany), and Solvay, Abbott,
primary prevention (two specifically involving and Mylan provided the n−3 fatty acid and pla-
patients with diabetes5,6 and two in broader popu- cebo capsules and some funding for packaging.
lations7,8) have been reported; none showed a clear Bayer (Germany) commented on the design of the
benefit or reported detailed information regarding trial, and both Bayer and Mylan commented on
bleeding events, so the balance of benefits and the draft of the manuscript but had no part in the
risks associated with aspirin use for primary pre- collection, handling, analysis, or interpretation of
vention among persons with diabetes remains the data or in the decision to submit the manu-
uncertain. Partly as a result of these studies, there script for publication. The manuscript was pre-
has been speculation that diabetes may be as- pared by the writing committee and reviewed and
sociated with reduced efficacy of the antiplatelet approved for submission for publication by the
effects of aspirin.9 steering committee. The first and last members
Retrospective meta-analyses of selected ran- of the writing committee vouch for the accuracy
domized trials of mainly low-dose aspirin have and completeness of the data and analyses and
suggested that aspirin use may result in an inci- for the fidelity of the trial to the protocol and the
dence of cancer or death from cancer that is 15 to data analysis plan.
20% lower than that with control.10-13 In particu-
lar, reductions of 30 to 40% in the incidence of Data Sharing
gastrointestinal cancers (particularly colorectal Deidentified data about the individual participants
cancer) were observed, with the effects appearing are to be shared with the Antithrombotic Trialists’
to increase with more prolonged exposure and Collaboration for meta-analysis. Requests for data
with longer follow-up up to 20 years. Data from sharing will be handled in line with the data ac-
randomized trials of sufficient size and duration cess and sharing policy of the Nuffield Depart-

2 n engl j med nejm.org

The New England Journal of Medicine

Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 Aspirin for Primary Prevention in Diabetes Mellitus

ment of Population Health, University of Oxford in discontinuation of the trial regimen, and any
(www​.­ndph​.­ox​.­ac​.­uk/​­about/​­data​-­access​-­policy). symptomatic bleeding episodes for which patients
saw a doctor. After a mean follow-up of 2.5 years,
Participants we requested blood and urine samples, along with
Men and women at least 40 years of age were measures of blood pressure and weight, from
considered to be eligible if they had received a 1800 randomly selected participants. (Details are
diagnosis of diabetes mellitus (any type) and did provided in the Methods section in Supplemen-
not have known cardiovascular disease and if tary Appendix 1, available at NEJM.org.)
there was substantial uncertainty about whether
antiplatelet therapy would confer worthwhile ben- Outcomes
efit. Key exclusion criteria were a clear indication While recruitment was still ongoing, we modified
for aspirin or a contraindication to aspirin or the the original primary outcome to include transient
presence of other clinically significant conditions ischemic attack in the definition of serious vascu-
that might limit adherence to the trial regimen lar event, a change that was made to increase the
for at least 5 years. All the participants provided statistical power of the trial. Thus, the prespeci-
written informed consent. fied primary efficacy outcome was the first seri-
ous vascular event, which was defined as a com-
Procedures posite of nonfatal myocardial infarction, nonfatal
Potential participants were identified from re- stroke (excluding confirmed intracranial hemor-
gional diabetes registers or general practices from rhage) or transient ischemic attack, or death from
around the United Kingdom and were sent a any vascular cause (excluding confirmed intracra-
screening questionnaire. Those who returned the nial hemorrhage). The primary safety outcome
questionnaire indicating that they were willing was the first occurrence of any major bleeding
and eligible to participate entered a prerandom- event, which was defined as a composite of any
ization run-in phase, during which placebo aspi- confirmed intracranial hemorrhage, sight-threat-
rin (and placebo n−3 fatty acids) was supplied. ening bleeding event in the eye, gastrointestinal
Their family doctor was informed of their poten- bleeding, or any other serious bleeding (i.e., a
tial participation, and they were sent a kit to ob- bleeding event that resulted in hospitalization or
tain blood and urine samples and to record blood transfusion or that was fatal). Secondary outcomes
pressure, height, and weight. After this run-in were gastrointestinal tract cancer (overall and ex-
period of 8 to 10 weeks, participants remained cluding those occurring during the first 3 years of
eligible if they returned a randomization ques- follow-up) and the composite of any serious
tionnaire confirming their willingness to continue, vascular event or any arterial revascularization
they still met the eligibility criteria, and they had procedure.
adhered to the trial regimen. All the reports of possible primary or second-
Using minimized randomization, we then as- ary outcomes were adjudicated centrally by clini-
signed eligible participants to receive 100 mg of cians who were unaware of the trial-group assign-
aspirin once daily or a matching placebo tablet16; ments in accordance with prespecified definitions.
participants were also assigned to receive 1-g cap- The data analysis plan prespecified that analyses
sules containing n−3 fatty acid once daily or a would be based on all the confirmed events plus
matching placebo capsule. The participants were unrefuted events (see the Methods section in Sup-
then mailed a 6-month supply of aspirin or placebo plementary Appendix 1).
tablets and n−3 fatty acids or placebo capsules,
as appropriate. Statistical Analysis
After randomization, we sent follow-up ques- The data analysis plan was finalized by the steer-
tionnaires and appropriate tablets and capsules ing committee and was published16 while all the
to participants every 6 months until the end of the members were still unaware of the trial results
trial. In these questionnaires, we sought informa- according to group assignment (except for the
tion regarding all serious adverse events (including statistician, who was aware of these assignments
potential trial outcomes), adherence to the trial and who was not involved in development of the
regimen, use of nontrial antiplatelet or anticoagu- data analysis plan). In addition to revising the
lant therapy, nonserious adverse events resulting definition of serious vascular events to include

n engl j med nejm.org 3
The New England Journal of Medicine
Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

transient ischemic attack, the sample was in- R e sult s

creased to at least 15,000 participants and the
duration of follow-up was increased to at least Trial Participants
7 years to increase the power of the trial. (Details From June 2005 through July 2011, a total of
are provided in the Methods section in Supple- 15,480 participants underwent randomization.
mentary Appendix 1.) The main prognostic characteristics of the par-
On the basis of an event rate of 1.2 to 1.3% ticipants were well balanced between the ran-
per year, as observed in both groups combined domized groups (Table 1, and Tables S1 and S2
when recruitment was complete, we determined in Supplementary Appendix 1). Aspirin use be-
that 7.5 years of the scheduled trial regimen would fore screening was reported by 5508 participants
provide the trial with 90% power, at a P value of overall (35.6%), but the treating doctor and the
less than 0.05, to detect a 15% difference between participant were sufficiently uncertain about its
the treatment groups in the risk of a serious vas- value to agree to randomization between aspirin
cular event. The expected number of gastrointes- and placebo.
tinal tract cancers was estimated to provide the At the end of the scheduled follow-up period,
trial with 60% power to detect a 30% lower risk complete follow-up data were available for 15,341
in the aspirin group than in the placebo group participants (99.1%) who had undergone ran-
during this period, but the prespecified focus for domization (Fig. S1 and Table S3 in Supplemen-
assessing the effects on cancer is 5 and 10 years tary Appendix 1). The mean follow-up was 7.4
after the end of the scheduled intervention phase. years, which yielded 57,000 person-years in the
We used log-rank methods to conduct inten- aspirin group and 56,945 person-years in the
tion-to-treat comparisons in time-to-event analy- placebo group. Adjudication was complete for
ses of the first occurrence of each type of event more than 90% of the primary and secondary
of interest among participants in the aspirin group outcomes.
as compared with those in the placebo group.17,18
A two-tailed P value of less than 0.05 was consid- Adherence and Effects on Vascular Risk
ered to indicate statistical significance for the Factors
primary efficacy and safety outcomes. It was The estimated mean adherence (weighted accord-
prespecified that the combined secondary out- ing to person-years at risk for a serious vascular
come of serious vascular event or revasculariza- event) to the assigned regimen was 70% in the
tion would be used for any subgroup analyses. aspirin group and 70% in the placebo group. Dur-
We made allowance for multiple hypothesis test- ing follow-up, adherence to the aspirin regimen
ing in the interpretation of secondary and explor- declined while the use of nontrial aspirin and
atory outcomes, with no formal adjustment to the other antiplatelet treatment increased. The esti-
P values. Consequently, the results are reported as mated mean between-group difference in the rate
point estimates and 95% confidence intervals of use of trial aspirin or nontrial aspirin or other
that have not been adjusted for multiple com- antiplatelet treatment was 69 percentage points
parisons, so the confidence intervals should (Table S4 in Supplementary Appendix 1). The
not be used to infer definitive treatment effects reasons for discontinuation of the trial regimen
within subgroups or with regard to secondary are shown in Table S5 in Supplementary Appen-
outcomes. dix 1. The use of nontrial medications (as reported
The baseline 5-year risk of vascular events at randomization and after a mean follow-up of
among participants was categorized into three 6.7 years) was similar in the two groups (Table S6
groups — less than 5%, 5% to less than 10%, in Supplementary Appendix 1).
and 10% or more — with the use of a risk score Table S7 in Supplementary Appendix 1 shows
that had been developed with the use of Poisson data regarding various vascular risk factors at a
regression. Details regarding this and other sec- mean of 2.5 years after randomization in the se-
ondary and exploratory assessments are provid- lected subgroup of approximately 1000 partici-
ed in the data analysis plan16 and in Supplemen- pants who were broadly representative of the trial
tary Appendix 1. The Clinical Trial Service Unit population. Findings were similar in the two trial
at the University of Oxford holds the full data- groups except for small differences in the systolic
base and performed all the analyses. blood pressure and the estimated glomerular fil-

4 n engl j med nejm.org

The New England Journal of Medicine

Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 Aspirin for Primary Prevention in Diabetes Mellitus

Table 1. Key Characteristics of the Participants at Baseline.*

Aspirin Group Placebo Group

Characteristic (N = 7740) (N = 7740)
Age
Mean — yr 63.2±9.2 63.3±9.2
Distribution — no. (%)
<60 yr 2795 (36.1) 2795 (36.1)
60 to <70 yr 3123 (40.3) 3124 (40.4)
≥70 yr 1822 (23.5) 1821 (23.5)
Male sex — no. (%) 4843 (62.6) 4841 (62.5)
White race — no. (%)† 7467 (96.5) 7468 (96.5)
Body-mass index‡
Mean 30.8±6.2 30.6±6.3
Distribution — no. (%)
<25 1080 (14.0) 1169 (15.1)
25 to <30 2753 (35.6) 2776 (35.9)
≥30 3665 (47.4) 3536 (45.7)
Unknown 242 (3.1) 259 (3.3)
Smoking status — no. (%)
Current smoker 639 (8.3) 640 (8.3)
Former smoker 3526 (45.6) 3525 (45.5)
Never smoked 3489 (45.1) 3488 (45.1)
Unknown 86 (1.1) 87 (1.1)
Participant-reported hypertension — no. (%) 4766 (61.6) 4767 (61.6)
Aspirin use before screening — no. (%) 2740 (35.4) 2768 (35.8)
Statin use — no. (%) 5854 (75.6) 5799 (74.9)
Type 2 diabetes — no. (%)§ 7282 (94.1) 7287 (94.1)
Duration of diabetes
Median (interquartile range) — yr 7 (3–13) 7 (3–13)
Distribution — no. (%)
<9 yr 4337 (56.0) 4322 (55.8)
≥9 yr 2976 (38.4) 2989 (38.6)
Unknown 427 (5.5) 429 (5.5)
Systolic blood pressure
Mean — mm Hg 136.1±15.2 136.2±15.3
Distribution — no. (%)
<130 mm Hg 1694 (21.9) 1700 (22.0)
≥130 to <140 mm Hg 1550 (20.0) 1541 (19.9)
≥140 mm Hg 2263 (29.2) 2292 (29.6)
Unknown 2233 (28.9) 2207 (28.5)
Vascular risk score — no. (%)¶
Low 3128 (40.4) 3136 (40.5)
Moderate 3294 (42.6) 3254 (42.0)
High 1318 (17.0) 1350 (17.4)

* Plus–minus values are means ±SD. Numbers and percentages are shown for categorical variables, and means or medians
(with interquartile ranges) for continuous variables. There were no significant differences between the assigned groups.
Percentages may not total 100 because of rounding.
† Race and ethnic group were reported by the participant. Other groups were Indian, Pakistani, or Bangladeshi (1% of
participants), African or Caribbean (1%), and other or unknown (1%).
‡ The body-mass index (the weight in kilograms divided by the square of the height in meters) was based on values for
height and weight that were reported by the participants.
§ The presence of type 2 diabetes was based on a broad clinical definition involving the age of the participant at the diag-
nosis of diabetes, the use of insulin within 1 year after diagnosis, and the body-mass index.
¶ We categorized the predicted 5-year risk of serious vascular event (including transient ischemic attack) without the use of aspi-
rin or n−3 fatty acids as follows: low risk as less than 5%, moderate risk as 5% to less than 10%, and high risk as 10% or more.

n engl j med nejm.org 5
The New England Journal of Medicine
Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

tration rate, which may well be chance findings Effects on Vascular Events and Bleeding
given the multiple comparisons. According to Baseline Characteristics
In exploratory analyses, the proportional effects
Effects on the Primary and Secondary of aspirin use on the combined outcome of seri-
Vascular Outcomes ous vascular events or revascularization and on
During the scheduled intervention period, the the safety outcome of major bleeding did not show
primary efficacy outcome occurred in a signifi- clear evidence of variation according to particular
cantly lower percentage of participants in the as- baseline characteristics (with allowance for multi-
pirin group than in the placebo group (658 ple comparisons). In particular, neither outcome
[8.5%] vs. 743 [9.6%]; rate ratio, 0.88; 95% CI, varied according to group assignment with re-
0.79 to 0.97; P = 0.01) (Fig. 1A). In exploratory gard to n−3 fatty acids or to the vascular risk
analyses, the risk difference was seen mainly in group at baseline (Figs. S4 and S5 in Supplemen-
the first 5 years, with no further gain subse- tary Appendix 1). Figure 3 shows that the inci-
quently in the number of participants avoiding dence of a major bleeding event increased with
an event (Fig. 1B). The effects on the components vascular risk. Substantial uncertainty around the
of the primary efficacy outcome and the second- observed number of events caused and avoided
ary outcome of serious vascular event or any results from small numbers and differences in
arterial revascularization are shown in Figure 2. adherence. Figure S6 in Supplementary Appendix 1
Prespecified exploratory analyses showed no sig- shows the predicted absolute effects with extrap-
nificant effect of aspirin use, as compared with olation to full adherence to the trial regimen.
placebo, on the rate of death from all vascular
causes combined, which represented approximate- Effects on Other Vascular
ly 30% of all deaths (Fig. S2 in Supplementary and Microvascular Outcomes
Appendix 1). Results of exploratory analyses of the effects of
aspirin use on other vascular and selected micro-
Effects on the Primary Safety Outcome vascular outcomes are shown in Figure S7 and
and Other Bleeding Table S9 in Supplementary Appendix 1. The re-
There was a significant adverse effect of assign- sults regarding the vascular events show trends
ment to the aspirin group, as compared with the that are generally similar to those regarding se-
placebo group, on the incidence of major bleeding rious vascular events. There was no apparent effect
(314 participants [4.1%] vs. 245 [3.2%]; rate ratio, of aspirin use on selected microvascular events.
1.29; 95% CI, 1.09 to 1.52; P = 0.003) (Fig. 2). Ex-
ploratory analyses did not suggest an attenua- Effects on Cancer and Other Nonvascular
tion of the effect on bleeding over time (Fig. S3 Outcomes
in Supplementary Appendix 1). Of the first major There was no between-group difference in the
bleeding events, 41.3% were gastrointestinal (of risk of gastrointestinal tract cancer, nor was there
which 62.3% were in the upper gastrointestinal a suggestion of an effect emerging with longer
tract, 32.9% were in the lower gastrointestinal follow-up (Fig. S8 in Supplementary Appendix 1).
tract, 2.2% were perforations, and 2.6% were The trial groups also did not differ significantly
undetermined), 21.1% were sight-threatening with regard to the risk of fatal or nonfatal cancer
bleeding events in the eye, 17.2% were intracra- overall or at particular sites (Table 2, and Fig. S9
nial bleeding events, and 20.4% were bleeding in Supplementary Appendix 1) or with regard to
events in other sites (mainly hematuria and epi- the risks of death overall or death from cancer
staxis that met the definition of major bleeding). or from all nonvascular causes (Fig. S2 in Sup-
The incidence of fatal bleeding events was simi- plementary Appendix 1). Searchable tabulations
lar among persons in the aspirin group and of the serious adverse events (fatal and nonfatal
among those in the placebo group (19 partici- combined) are provided in Supplementary Appen-
pants [0.2%] and 16 [0.2%], respectively), as was dix 2, available at NEJM.org. They are grouped on
the incidence of hemorrhagic stroke (25 [0.3%] the basis of the Medical Dictionary for Regulatory
and 26 [0.3%]) (Table S8 in Supplementary Ap- Activities, version 14.0, classification system, ac-
pendix 1). cording to system organ class.

6 n engl j med nejm.org

The New England Journal of Medicine

Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 Aspirin for Primary Prevention in Diabetes Mellitus

Discussion ment to the use of aspirin at a dose of 100 mg

daily for 7.4 years resulted in a risk of serious
In this trial involving persons who had diabetes vascular events that was 12% lower than that with
without manifest cardiovascular disease, assign- placebo but also in a risk of major bleeding that

A First Serious Vascular Event

100 15
90 Rate ratio, 0.88 (95% CI, 0.79–0.97)
P=0.01 Placebo
80
Participants with Event (%)

10
70 Aspirin
60
5
50
40
30 0
0 1 2 3 4 5 6 7 8 9
20
10
0
0 1 2 3 4 5 6 7 8 9
Years of Follow-up
No. at Risk
Placebo 7740 7618 7486 7342 7188 7001 5771 3890 2200 1430
Aspirin 7740 7655 7536 7404 7252 7096 5825 3966 2222 1428

Cumulative benefit per 1000 4±2 6±2 9±3 10±3 13±4 11±4 12±5 9±6 10±7
participants in aspirin group

B First Serious Vascular Event, According to Year of Follow-up

Year of First Aspirin Placebo
Event (N=7740) (N=7740) Rate Ratio (95% CI) P Value
no. of participants with event (%)
<3 201 (2.6) 269 (3.5) 0.74 (0.62–0.89)
3 to <5 169 (2.3) 198 (2.7) 0.85 (0.69–1.04)
≥5 to <7 180 (2.5) 171 (2.4) 1.04 (0.84–1.28)
7 108 (2.7) 105 (2.7) 1.02 (0.78–1.33)
All 658 (8.5) 743 (9.6) 0.88 (0.79–0.97) 0.01
0.6 0.8 1.0 1.2 1.4 1.6

Aspirin Better Placebo Better

Figure 1. First Serious Vascular Event during Follow-up.

Panel A shows a Kaplan–Meier plot of the first serious vascular event (a composite of nonfatal myocardial infarc-
tion, nonfatal ischemic stroke or transient ischemic attack, or death from any vascular cause, excluding confirmed
intracranial hemorrhage) during follow-up. The numbers of participants at risk at the start of each year of follow-up
are shown, along with the cumulative number (±SE) of participants per 1000 in the aspirin group as compared with
the placebo group who avoided events. The inset shows the same data on an expanded y axis. Panel B shows the
rate ratios for the first serious vascular event among the participants in the aspirin group, as compared with those
in the placebo group, according to the period of follow-up. The numbers at risk declined with each period of follow-
up because of censoring, so the percentages are the number of events as a proportion of the number at risk at the
start of each period. For each period of follow-up, rate ratios are plotted as squares, with the size of each square
proportional to the amount of statistical information that was available; the horizontal lines represent 95% confi-
dence intervals (not adjusted for multiple comparisons), and the dashed vertical line indicates the overall rate ratio
for the effect of aspirin use on the first serious vascular event. For the prespecified composite period of follow-up,
the rate ratio and corresponding 95% confidence interval are represented by a diamond. Squares or a diamond to
the left of the solid vertical line indicate a benefit with aspirin use, but the comparison was significant (P<0.05) only
if the horizontal line or diamond does not overlap with the solid vertical line. The test for trend across years was sig-
nificant (χ 2 = 6.24; P = 0.01).

n engl j med nejm.org 7
The New England Journal of Medicine
Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Aspirin Placebo
Type of Event (N=7740) (N=7740) Rate Ratio (95% CI) P Value
no. of participants with event (%)
Vascular Outcomes
Nonfatal myocardial infarction 191 (2.5) 195 (2.5) 0.98 (0.80–1.19)
Nonfatal presumed ischemic stroke 202 (2.6) 229 (3.0) 0.88 (0.73–1.06)
Vascular death excluding intracranial hemorrhage 197 (2.5) 217 (2.8) 0.91 (0.75–1.10)
Any serious vascular event excluding TIA 542 (7.0) 587 (7.6) 0.92 (0.82–1.03)
TIA 168 (2.2) 197 (2.5) 0.85 (0.69–1.04)
Any serious vascular event including TIA 658 (8.5) 743 (9.6) 0.88 (0.79–0.97) 0.01
Any arterial revascularization 340 (4.4) 384 (5.0) 0.88 (0.76–1.02)
Any serious vascular event or revascularization 833 (10.8) 936 (12.1) 0.88 (0.80–0.97)

Major Bleeding
Intracranial hemorrhage 55 (0.7) 45 (0.6) 1.22 (0.82–1.81)
Sight-threatening bleeding in eye 57 (0.7) 64 (0.8) 0.89 (0.62–1.27)
Serious gastrointestinal bleeding 137 (1.8) 101 (1.3) 1.36 (1.05–1.75)
Other major bleeding 74 (1.0) 43 (0.6) 1.70 (1.18–2.44)
Any major bleeding 314 (4.1) 245 (3.2) 1.29 (1.09–1.52) 0.003
0.5
0.7 1.0 1.5 2.0

Aspirin Better Placebo Better

Figure 2. Effect of Assignment to Aspirin Group on Components of Serious Vascular Events, the Combined Outcome of Serious Vascular
Event or Revascularization, and Major Bleeding and Its Components.
The primary outcome was a serious vascular event (a composite of nonfatal myocardial infarction, nonfatal ischemic stroke or transient
ischemic attack [TIA], or death from any vascular cause, excluding confirmed intracranial hemorrhage). Secondary outcomes were a seri-
ous vascular event or any coronary or noncoronary revascularization procedure. A single participant may have had multiple events and
therefore may contribute information to more than one row. The size of each square for the rate ratio is proportional to the amount of
statistical information that was available, the horizontal lines represent 95% confidence intervals, and the dashed vertical line indicates
the overall rate ratio for the effect of aspirin use on the first serious vascular event. An arrow on the horizontal line indicates that the
confidence interval exceeds the graph area. For composite outcomes, rate ratios and their corresponding 95% confidence intervals are
represented by diamonds. Bold entries with diamonds show totals for all data listed above them. The effect of aspirin use on the com-
ponents of the primary safety outcome of major bleeding event (a composite of intracranial hemorrhage, sight-threatening bleeding
event in the eye, gastrointestinal bleeding, or other major bleeding event) is shown with the use of the same graphic conventions.

was 29% higher. The lower risk of serious vascu- of the participants who underwent randomiza-
lar events is similar to the risk that was reported tion have allowed reliable detection of these
previously in the Antithrombotic Trialists’ Col- moderate but important effects on the incidence
laboration meta-analysis of primary prevention of vascular events and on both the severity and
trials of similar doses of aspirin (which used incidence of bleeding. Our findings do not sup-
slightly different outcome definitions; see the port the hypothesis that persons with diabetes
Methods section in Supplementary Appendix 1).1 have a resistance to aspirin.9 Although the pro-
In contrast to those previous trials, there were portional effects of aspirin use are likely to be
high rates of the use of cardioprotective treat- generalizable to the wider population of persons
ments among the participants in ASCEND, with with diabetes, the absolute event rates and ad-
the majority of participants taking statins and herence rates reflect this population of persons
blood pressure–lowering therapy. Hence, the pres- with well-treated diabetes. Overall, on the basis
ent trial provides a direct assessment of the bal- of the absolute percentage differences between
ance of the benefits and hazards of aspirin use the groups in the incidence of serious vascular
in a contemporary context. events (1.1 percentage points lower in the aspi-
In our trial, factors such as the large number rin group than in the placebo group) and in
of participants and clinical outcomes, long dura- bleeding events (0.9 percentage points higher in
tion of follow-up, the randomized, blinded design the aspirin group), 91 patients would need to be
of the trial, and the almost complete follow-up treated to avoid a serious vascular event over a

8 n engl j med nejm.org

The New England Journal of Medicine

Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 Aspirin for Primary Prevention in Diabetes Mellitus

200 Nonfatal myocardial infarction, Nonfatal major bleeding

nonfatal stroke, transient ischemic Fatal bleeding
attack, or revascularization 176.4
180
Death from cardiovascular cause
165.1
160

140

Events per 5000 Person-Yr

120

100 96.9

83.5
80

60 54.3
44.0 44.7
40 37.9
28.6
18.1 15.2 19.8
20

0
rin

rin

rin

rin

rin

rin

o
eb

eb

eb

eb

eb

eb
pi

pi

pi

pi

pi

pi
ac

ac

ac

ac

ac

ac
As

As

As

As

As

As
Pl

Pl

Pl

Pl

Pl

Pl
Serious Major Serious Major Serious Major
Vascular Bleeding Vascular Bleeding Vascular Bleeding
Event or Event or Event or
Revascular- Revascular- Revascular-
ization ization ization
<5% 5% to <10% ≥10%
(40.5% of trial population) (42.3% of trial population) (17.2% of trial population)
Estimated 5-Yr Risk of Serious Vascular Event at Baseline
No. of Events per 5000 Person-Yr in Aspirin Group
Serious vascular events avoided 5.7±3.7 11.2±5.4 4.9±12.9
Serious vascular events or revascularizations avoided 6.1±4.2 13.4±6.3 11.3±14.3
Major bleeding caused 2.8±2.6 8.9±3.2 9.6±7.5

Figure 3. Observed Absolute Effect of Assignment to Aspirin Group on Serious Vascular Events or Revascularization and on Major
Bleeding, According to Vascular Risk.
Shown is the observed absolute incidence of serious vascular events or revascularization and of major bleeding events in the aspirin
group, as compared with the placebo group, according to the three baseline levels of vascular risk (predicted 5-year risks of a serious
vascular event, with low risk defined as <5%, moderate risk as 5% to <10%, and high risk as ≥10%), with data expressed as numbers
of events per 5000 person-years. Plus–minus values are means ±SE.

period of 7.4 years, and 112 to cause a major incidence of serious vascular events and bleed-
bleeding event. ing were the same across different levels of vas-
These results of intention-to-treat analyses cular risk. The results of these analyses provide
tend to underestimate the effect of actual aspirin a more reliable estimate of the absolute differ-
use, both with respect to events avoided and ences in the event rates resulting from aspirin
bleeding events caused owing to a lack of full use by applying the extrapolated overall rate ra-
adherence to the regimen during the trial. Explor- tios to the event rates with placebo in each risk
atory analyses comparing participants at different group. On the basis of these assumptions, the
levels of vascular risk extrapolated the rate ratios predicted number of serious vascular events that
in the intention-to-treat analysis to full adherence would be avoided by participants actually taking
and assumed that the proportional effects on the aspirin was closely balanced by the predicted

n engl j med nejm.org 9
The New England Journal of Medicine
Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 2. Effect of Aspirin Use on the Incidence of Site-Specific Fatal or Nonfatal Cancer.*

Aspirin Group Placebo Group

Cancer Type (N = 7740) (N = 7740) Rate Ratio (95% CI)

no. of participants (%)

Gastrointestinal tract cancer 157 (2.0) 158 (2.0) 0.99 (0.80–1.24)
Other gastrointestinal cancer† 87 (1.1) 82 (1.1) 1.06 (0.78–1.43)
Respiratory cancer 101 (1.3) 103 (1.3) 0.98 (0.74–1.29)
Genitourinary cancer 332 (4.3) 294 (3.8) 1.13 (0.97–1.32)
Hematologic cancer 88 (1.1) 86 (1.1) 1.02 (0.76–1.38)
Breast cancer 97 (1.3) 96 (1.2) 1.01 (0.76–1.34)
Melanoma 50 (0.6) 59 (0.8) 0.85 (0.58–1.23)
Other cancer 25 (0.3) 30 (0.4) 0.83 (0.49–1.41)
Unspecified cancer 26 (0.3) 31 (0.4) 0.84 (0.50–1.41)
Any cancer‡ 897 (11.6) 887 (11.5) 1.01 (0.92–1.11)

* The 95% confidence intervals were unadjusted for multiple comparisons. A single participant may have had multiple cancers.
† Other gastrointestinal cancer includes hepatobiliary and pancreatic cancers.
‡ Any cancer excludes nonfatal nonmelanoma skin cancer and nonfatal recurrence of a cancer that had occurred before
randomization.

number of major bleeding events caused, even per gastrointestinal tract that has been seen in
among persons who had a 5-year vascular risk of short-term studies,23 as well as confirming long-
10% or more (Fig. S6 in Supplementary Appen- term safety.
dix 1). A recent analysis suggesting a greater Several meta-analyses of selected randomized
benefit of low-dose aspirin use on the incidence trials of generally low-dose aspirin have suggest-
of vascular events among persons with a body ed that aspirin use might reduce the risk of cancer
weight of less than 70 kg was not confirmed in — in particular, gastrointestinal tract cancer — by
exploratory subgroup analyses (and, indeed, a up to one third during long-term follow-up, with
trend in the opposite direction was observed).19 effects becoming apparent approximately 3 years
The assessment of the balance between the after randomization.10,12,13,24 However, despite more
benefit and harm of aspirin use in the context of than 7 years of aspirin treatment and follow-up
primary prevention is complicated by the difficulty in ASCEND, we found no evidence of a reduction
of comparing the severity of the vascular events in the incidence of gastrointestinal tract cancer or
avoided and the bleeding events caused. For ex- of cancer at any other site, even during the later
ample, although transient ischemic attacks are years of follow-up. These analyses had limited sta-
minor in themselves, they are associated with in- tistical power to detect the hypothesized effects,
creased risks of subsequent stroke and cognitive so follow-up is being continued through central
impairment.20 Approximately half the excess of registries.
bleeding was in the gastrointestinal tract, with In conclusion, the use of low-dose aspirin led
approximately one third in the upper gastroin- to a lower risk of serious vascular events than
testinal tract. However, even near the end of the placebo among persons with diabetes who did
trial in 2016, only approximately one quarter of not have evident cardiovascular disease at trial
participants were receiving proton-pump inhibi- entry. However, the absolute lower rates of seri-
tors (PPIs). It is possible that bleeding rates among ous vascular events were of similar magnitude to
aspirin users might be lower if PPIs were rou- the absolute higher rates of major bleeding, even
tinely used in these persons, provided that longer- among participants who had a high vascular risk.
term trials of PPIs21,22 confirm the substantial The use of low-dose aspirin did not result in a
reductions in the incidence of bleeding in the up- lower risk of gastrointestinal tract cancer or other

10 n engl j med nejm.org

The New England Journal of Medicine

Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.
 Aspirin for Primary Prevention in Diabetes Mellitus

cancer over the mean follow-up of 7.4 years, but sity of Oxford), the British Heart Foundation, and Cancer Re-
search U.K.
further follow-up is needed to assess any longer- Disclosure forms provided by the authors are available with
term effects on cancer reliably. the full text of this article at NEJM.org.
Supported by grants to the University of Oxford from the Brit- A data sharing statement provided by the authors is available
ish Heart Foundation and by Bayer (Germany and the United with the full text of this article at NEJM.org.
States), Solvay, Abbott, and Mylan. The Clinical Trial Service We thank the participants, the collaborating doctors and gen-
Unit at the University of Oxford receives support from the U.K. eral practitioners, the members of the steering committee and
Medical Research Council (which funds the MRC Population the data monitoring committee, and all the administrative and
Health Research Unit in a strategic partnership with the Univer- clinical staff who helped to conduct the trial.

Appendix
The affiliations of the members of the writing committee are as follows: Clinical Trial Service Unit and Epidemiological Studies Unit,
Nuffield Department of Population Health (L.B., M.M., K.W., W.S., G.B., J. Barton, K.M., R.H., J.C., A.M., A.Y., M.L., F.C., E.S., E.W.,
D.S., R.P., C.B., R.C., S.P., J.A.), Nuffield Department of Primary Care Health Sciences (A.F.), and Wolfson College (A.N.), University
of Oxford, Oxford, the Royal Berkshire NHS Foundation Trust, Reading (T.A.), Addenbrooke’s Hospital, Cambridge (A.A.), Leeds
General Infirmary, Leeds (J. Bodansky), and University Hospital of Wales, Cardiff (R.M.) — all in the United Kingdom.

References
1. Antithrombotic Trialists’ Collabora- events in a general population screened for 17. Yusuf S, Peto R, Lewis J, Collins R,
tion. Aspirin in the primary and secondary a low ankle brachial index: a randomized Sleight P. Beta blockade during and after
prevention of vascular disease: collabora- controlled trial. JAMA 2010;​303:​841-8. myocardial infarction: an overview of the
tive meta-analysis of individual partici- 9. Ajjan R, Storey RF, Grant PJ. Aspirin randomized trials. Prog Cardiovasc Dis
pant data from randomised trials. Lancet resistance and diabetes mellitus. Diabeto- 1985;​27:​335-71.
2009;​373:​1849-60. logia 2008;​51:​385-90. 18. Peto R, Pike MC, Armitage P, et al.
2. Smith SC Jr, Benjamin EJ, Bonow RO, 10. Rothwell PM, Price JF, Fowkes FG, et Design and analysis of randomized clini-
et al. AHA/ACCF secondary prevention al. Short-term effects of daily aspirin on cal trials requiring prolonged observation
and risk reduction therapy for patients cancer incidence, mortality, and non-vas- of each patient. I. Introduction and de-
with coronary and other atherosclerotic cular death: analysis of the time course of sign. Br J Cancer 1976;​34:​585-612.
vascular disease: 2011 update: a guideline risks and benefits in 51 randomised con- 19. Rothwell PM, Cook NR, Gaziano JM,
from the American Heart Association and trolled trials. Lancet 2012;​379:​1602-12. et al. Effects of aspirin on risks of vascu-
American College of Cardiology Founda- 11. Rothwell PM, Wilson M, Elwin CE, et lar events and cancer according to body-
tion. Circulation 2011;​124:​2458-73. al. Long-term effect of aspirin on colorec- weight and dose: analysis of individual
3. The Emerging Risk Factors Collabora- tal cancer incidence and mortality: 20-year patient data from randomised trials. Lan-
tion. Diabetes mellitus, fasting blood glu- follow-up of five randomised trials. Lan- cet 2018;​392:​387-99.
cose concentration, and risk of vascular cet 2010;​376:​1741-50. 20. van Rooij FG, Kessels RPC, Richard E,
disease: a collaborative meta-analysis of 12. Rothwell PM, Wilson M, Price JF, De Leeuw F-E, van Dijk EJ. Cognitive im-
102 prospective studies. Lancet 2010;​375:​ Belch JF, Meade TW, Mehta Z. Effect of pairment in transient ischemic attack pa-
2215-22. daily aspirin on risk of cancer metastasis: tients: a systematic review. Cerebrovasc
4. IDF diabetes atlas. 8th ed. Brussels:​ a study of incident cancers during ran- Dis 2016;​42:​1-9.
International Diabetes Federation, 2017 domised controlled trials. Lancet 2012;​ 21. Bosch J, Eikelboom JW, Connolly SJ,
(http://www​.diabetesatlas​.org). 379:​1591-601. et al. Rationale, design and baseline char-
5. Belch J, MacCuish A, Campbell I, et al. 13. Flossmann E, Rothwell PM. Effect of acteristics of participants in the Cardio-
The Prevention of Progression of Arterial aspirin on long-term risk of colorectal vascular Outcomes for People Using Anti-
Disease and Diabetes (POPADAD) trial: cancer: consistent evidence from ran- coagulation Strategies (COMPASS) trial.
factorial randomised placebo controlled domised and observational studies. Lancet Can J Cardiol 2017;​33:​1027-35.
trial of aspirin and antioxidants in pa- 2007;​369:​1603-13. 22. Eikelboom JW, Connolly SJ, Bosch J,
tients with diabetes and asymptomatic 14. The ASCEND Study Collaborative et al. Rivaroxaban with or without aspirin
peripheral arterial disease. BMJ 2008;​337:​ Group. Effects of n−3 fatty acid supple- in stable cardiovascular disease. N Engl J
a1840. ments in diabetes mellitus. N Engl J Med. Med 2017;​377:​1319-30.
6. Ogawa H, Nakayama M, Morimoto T, DOI:​ 10.1056/NEJMoa1804989. 23. Scally B, Emberson JR, Spata E, et al.
et al. Low-dose aspirin for primary pre- 15. Aung T, Haynes R, Barton J, et al. Effects of gastroprotectant drugs for the
vention of atherosclerotic events in pa- Cost-effective recruitment methods for a prevention and treatment of peptic ulcer
tients with type 2 diabetes: a randomized large randomised trial in people with dia- disease and its complications: a meta-
controlled trial. JAMA 2008;​300:​2134-41. betes: A Study of Cardiovascular Events analysis of randomised trials. Lancet Gas-
7. Ikeda Y, Shimada K, Teramoto T, et al. iN Diabetes (ASCEND). Trials 2016;​17:​286. troenterol Hepatol 2018;​3:​231-41.
Low-dose aspirin for primary prevention 16. Bowman L, Mafham M, Stevens W, et 24. Rothwell PM, Fowkes FG, Belch JF,
of cardiovascular events in Japanese pa- al. ASCEND: A Study of Cardiovascular Ogawa H, Warlow CP, Meade TW. Effect
tients 60 years or older with atheroscle- Events iN Diabetes: characteristics of a of daily aspirin on long-term risk of death
rotic risk factors: a randomized clinical randomized trial of aspirin and of ome- due to cancer: analysis of individual pa-
trial. JAMA 2014;​312:​2510-20. ga-3 fatty acid supplementation in 15,480 tient data from randomised trials. Lancet
8. Fowkes FG, Price JF, Stewart MC, et al. people with diabetes. Am Heart J 2018;​ 2011;​377:​31-41.
Aspirin for prevention of cardiovascular 198:​135-44. Copyright © 2018 Massachusetts Medical Society.

n engl j med nejm.org 11
The New England Journal of Medicine
Downloaded from nejm.org on August 26, 2018. For personal use only. No other uses without permission.
Copyright © 2018 Massachusetts Medical Society. All rights reserved.