European Journal of Cancer (2013) 49, 1049– 1057

A v a i l a b l e a t w w w . s c i e n c e d i r e c t . c o m

jour na l homepage : www.e jcancer . in fo

Use of aspirin post-diagnosis in a cohort of patientswith colorectal cancer and its association with all-causeand colorectal cancer specific mortality

C. McCowan ⇑, A.J. Munro, P.T. Donnan, R.J.C. Steele

Division of Population Health Sciences, Medical Research Institute, University of Dundee, Dundee DD2 4BF, United Kingdom

Available online 19 November 2012

09

ht

⇑+

D

KEYWORDS

AspirinColorectal cancerSurvival

59-8049/$ - see front matter

tp://dx.doi.org/10.1016/j.ejca.

Corresponding author: Add44 (0)1382 420145; fax: +44

E-mail addresses: c.mccowonnan), r.j.c.steele@dundee.a

� 2012 E

2012.10.0

ress: Mac(0)1382 4an@dunc.uk (R.J

Abstract Objective: Aspirin is associated with a reduced risk of developing colorectal cancer.This study examined whether patients with colorectal cancer prescribed aspirin had improvedsurvival.Design: An observational population cohort study was undertaken using data linkage of can-cer registry, dispensed prescriptions and death certificate records in Tayside, Scotland. Allcommunity prescribed aspirin pre- and post-diagnosis was extracted and periods of aspirinuse post-diagnosis for each individual were analysed using Cox proportional hazard models.Main outcome measures were all-cause and colorectal mortality from death certificates.Results: Two thousand nine hundred ninety patients were identified with colorectal cancerbetween 1st January 1997 and 30th December 2006 and followed up until 28th February 2010.Median age at diagnosis was 73 (interquartile range [IQR] 65–80) with 52% male. One thou-sand nine hundred ninety-eight (67%) deaths were recorded with 1021 (34%) attributed tocolorectal cancer. One thousand three hundred forty (45%) patients used aspirin at some stageof the study period.Aspirin use post-diagnosis was associated with lower risk of all cause mortality (hazard ratio[HR] = 0.67, 95% confidence interval [CI] = 0.57–0.79, p < 0.001) and colorectal cancer spe-cific mortality after allowing for age, Dukes’ stage, gender, socio-economic status and aspirinuse pre-diagnosis. Increasing age and stage at diagnosis were associated with increased risk,with more affluent patients at reduced risk.Conclusions: Our study suggests that aspirin use post-diagnosis of colorectal cancer mayreduce both all cause and colorectal cancer specific mortality. However further work isrequired to ensure this is a causal relationship and to identify whether it is best used in specificgroups of patients.� 2012 Elsevier Ltd. All rights reserved.

lsevier Ltd. All rights reserved.

24

kenzie Building, University of Dundee, Kirsty Semple Way, Dundee DD2 4BF, United Kingdom. Tel.:20010.dee.ac.uk (C. McCowan), a.j.munro@dundee.ac.uk (A.J. Munro), p.t.donnan@dundee.ac.uk (P.T..C. Steele).

1050 C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057

1. Introduction

There is now good evidence from randomised con-trolled trials that aspirin1–5 and other prostaglandin-endoperoxide synthase (PTGS, previously COX)inhibitors6–8 can reduce the recurrence of colorectaladenomas by around 20%. Given that most colorectalcancers arise from adenomas9 it follows that aspirinmight reduce the risk of developing colorectal cancerand this does seem to be the case. Long term follow upof two large trials of high dose (greater than 500 mgper day) aspirin designed for the prevention of cardiovas-cular events has demonstrated reduced risk of colorectalcancer after 10 years.10 More recently a follow-up of fourrandomised controlled trials of aspirin versus control inthe primary and secondary prevention of vascular eventsand one trial comparing different doses of aspirin hasshown a reduction in the 20 year risk of developing colo-rectal cancer (hazard ratio [HR] 0.76, 95% confidenceinterval [CI] 0.60–0.98) and disease specific mortality(HR 0.65, 95% CI 0.48–0.88).11

Interestingly, although the benefit increased withduration of treatment, no increase in benefit was seenat doses of aspirin greater than 75 mg per day. In addi-tion the colorectal adenoma/carcinoma prevention pro-gramme 2 (CAPP 2) study which was designed to look ataspirin and resistance starch in the prevention of cancerin hereditary nonpolyposis colorectal cancer (HNPCC)patients has, after prolonged follow-up, shown a signif-icant effect of aspirin in preventing the development ofcancer in these individuals.12

The reason for these observations is not clear but theconsistency of the evidence strongly indicates that aspi-rin has a truly preventative effect and the weight of evi-dence suggests that the effect may be mediated by theinhibition of PTGS.13,14 As PTGS 2 is over expressedin the majority of human colorectal cancers15,16 and isknown to mediate cell proliferation17 it is reasonableto hypothesise that PTGS 2 inhibition may have anti-tumour effects in established cancer. This is supportedby the finding that in animal models, drugs that inhibitPTGS 2 can also inhibit tumour growth and metasta-ses.18–24 Other work showing that aspirin also has aneffect on PGTS-1 in animals and that in humans poly-morphisms of this gene are associated with survivalmean that the effect of aspirin PTGS 1 and 2 may notbe completely understood.25,26

There is also supporting evidence from human stud-ies; in a study of adjuvant chemotherapy in stage 3 coloncancer patients post-diagnosis aspirin therapy wasfound to be associated with a lower risk of recurrenceor death.27 Furthermore in a prospective cohort studyof 1279 patients with non metastatic colorectal cancerderived from the Nurses’ Health Study and the HealthProfessional Follow-up Study the influence of pre- andpost-diagnosis aspirin use on survival has recently been

studied.28 This has demonstrated a multivariate hazardratio of 0.71 (95% CI 0.53–0.95) for colorectal cancerspecific mortality for regular aspirin users after diagno-sis and when those taking aspirin before diagnosis wereexcluded the HR reduced to 0.53 (95% CI 0.33–0.80). Asecond more recent study examined patients diagnosedwith colorectal cancer in a region of The Netherlandslinked to prescribing information and reported on all-cause mortality.29 They reported a multivariate hazardratio for post-diagnosis aspirin use in colon cancerpatients of 0.65 (95% CI 0.50–0.84) but no effect in rectalcancer patients (HR = 1.10 0.79–1.154). These studiesshow similar effects but one used an occupationallyselected group of patients and reported on cancer spe-cific outcomes whilst the second only showed an effectin one specific group of colorectal patients and reportedon all-cause mortality. We have conducted a popula-tion-based study to address the question – does use ofaspirin influence all-cause and colorectal specific mortal-ity in a geographically defined cohort of patients withcolorectal cancer.

2. Aims

To examine whether patients in Tayside, UnitedKingdom (UK) (population 396,000) diagnosed withcolorectal cancer, and subsequently prescribed aspirin,have improved survival compared to patients with colo-rectal cancer who have no documented use of aspirin.

3. Methods

The Health Informatics Centre (HIC, http://www.dundee.ac.uk/HIC) holds a number of healthrelated databases on the residents of Tayside, Scotland,UK. HIC contains data on every patient registered witha GP in the Tayside area. Each GP registered patient hasa unique 10 digit identifier called the community healthindex number (CHI). The CHI allows data linkage of allNHS encounters a person may have allowing the combi-nation of dispensed prescriptions with clinical contactsand outcomes.30 Dispensed prescribing relates to medi-cation which were prescribed to the patient by a doctorwhich have then been collected at a pharmacy.

We examined all cancer registry records in Taysideregion of Scotland for the period 1st January 1997 and31st December 2006 with any case of colorectal cancerextracted. Colorectal cancer was defined as a code relat-ing to ICD10 classifications C18 (Malignant neoplasmof colon), C19 (Malignant neoplasms of rectosigmoidjunction), C20 (Malignant neoplasm of rectum) andC21 (Malignant neoplasm of anus and anal canal). Indi-vidual patients were identified and linked to dispensedprescribing and death certificate records from 1st Janu-ary 1993 to 28th February 2010. Patients were classed as

C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057 1051

dying from colorectal cancer if this was listed on thedeath certificate as the underlying cause of death. Dis-pensed prescribing records were examined and all pre-scriptions relating to aspirin extracted.

Patients were initially identified as using aspirin pre-or post-diagnosis. All aspirin prescribing post-diagnosispatients were examined and the possible coverage ofeach prescription was calculated based on the numberof tablets and the directions, e.g. 28 tablets at one perday gave coverage for that prescription of 28 days.Patient follow-up post-diagnosis was split into periodsof aspirin use, or non-use, based on this coverage foreach individual until death or the end of follow-up.The total duration of aspirin use was calculated fromthe date of the first prescription post-diagnosis to theend of coverage of the last prescription. Dividing thetotal coverage of all prescriptions by the duration gavea measure of adherence over the period of aspirin use.30

The study was approved by the Tayside Committeeon Medical Research Ethics and the NHS Tayside Cal-dicott Guardian.

4. Statistical analysis

Data were described as number of subjects (percent-ages) for categorical variables and mean (s.d.) for con-tinuous variables. Where continuous variables did notfollow a normal distribution, they were tested usingthe Shapiro–Wilks test for skewness, and the medianand interquartile range (IQR) were reported. Likelihoodratios and chi-squared tests for trend (v2 trend, df, p)were reported for differences in distribution of the pop-ulation with n ordered categories, otherwise Pearson’sChi-squared test for differences was used (v2, df, p).

Cox proportional hazard models were utilised to esti-mate hazard ratios and 95% confidence intervals foreach unadjusted and adjusted covariate for cancer diag-nosis. Patients were followed up from cancer diagnosisuntil date of death or the end of the study. Individualpatients who used aspirin post-diagnosis had severallinked records each reflecting a period of aspirin use,or no use, which provided continuous follow-up. Thismarker of aspirin use was included in Cox proportionalhazard models to allow for the effect over time sincediagnosis to be accurately examined. The proportionalhazard assumption was assessed using trend-tests ofthe Schoenfeld residuals. The multiple regression analy-sis allowed for age, sex, socio-economic status, cancerstage at diagnosis and aspirin use pre-diagnosis.

All statistical analyses were performed using Stataversion 10.

5. Results

There were 3067 patients identified from CancerRegistry records as having an incident colorectal cancer

(ICD10 code C18–C21) between 1st January 1997 and31st December 2006, with all patients followed up untildeath or the end of the study period, 28th February2008. Thirty-five patients were excluded as they hadprevious colorectal cancer and 42 were excluded as theywere diagnosed at death leaving a total of 2990.

The median age at diagnosis was 73 (IQR 65–80) with52% of recorded cases in men. Patients were followed uppost-diagnosis for 11,393 patient years in total. Mortal-ity was high with 1998 (67%) deaths recorded during fol-low up, with a median follow-up for all patients of2.80 years (IQR = 0.63–6.21). There were 2754 (92%)patients followed up until death or for a minimum of5 years post-diagnosis. Median follow-up for survivingpatients was 7.13 years (IQR = 5.13–9.70). Colorectalcancer was the underlying cause of death on the deathcertificate for 1021 (34%) patients.

Using chi-squared tests to examine unadjusted differ-ences in all-cause mortality, there was a higher propor-tion of deaths in patients with cancer of therectosigmoid junction (C19) than other cancers(v2 = 14.91, df = 3, p = 0.002), in the least affluentpatients (v2 trend = 26.52, df = 4, p < 0.001), in olderpatients (v2 trend = 309, df = 4, p < 0.001) and thosewith more advanced Dukes’ Stage (v2 trend = 523,df = 3, p < 0.001) as shown in Table 1. There was no dif-ference by gender (v2 = 2.28, df = 1, p = 0.131).

There were 1340 (45%) patients who used aspirin atsome stage of the study period. Users of aspirin weremore likely to be older (v2 trend = 196.76, df = 4,p < 0.001) and to present with lower stage of disease(v2 trend = 47.44, df = 4, p < 0.001). However therewere no differences by site of cancer (v2 = 4.88, df = 3,p = 0.181), gender (v2 = 3.01, df = 1, p = 0.083) orsocial class (v2 trend = 2.24, df = 4, p = 0.692).

Combining the figures shown in Table 2 for use ofaspirin pre- and post-diagnosis and post-diagnosis only,there were a total of 894 (30%) patients who used aspirinpost-diagnosis with a median of 1.53 years use (IQR0.46–3.99 years). Calculated adherence to aspirin overthe entire period was high with a median figure of95.4% (IQR 84.3–100%). Out of this 859 (96.1%)patients only received aspirin at a 75 mg dose, 14(1.6%) had only 300 mg dose and 21 (2.3%) had a com-bination of the two. Overall, there were 19,422 prescrip-tions for aspirin post-diagnosis of cancer with 97.9% fora 75 mg dose. Other characteristics of patients by aspirinuse pre- and post-diagnosis are shown in Table 2.

Cox proportional hazard models for all-cause mortal-ity are reported unadjusted for age, sex, social class,stage of disease at diagnosis and aspirin use (see Table 3).A multiple regression model was also created andshowed aspirin use post-diagnosis to be associated withlower risk of mortality (HR = 0.67, 95% CI = 0.57–0.79, p < 0.001) after adjusting for other covariates.Increasing age and stage at diagnosis were associated

Table 2Characteristics of the cohort by aspirin use.

None(%)

Aspirin use

Pre-diagnosisonly (%)

Pre- & post-diagnosis (%)

Post-diagnosisonly (%)

Patients (%) 1650(55)

446 (15) 544 (18) 350 (12)

ICD10 classification

C18 (colon) 801(49)

242 (54) 243 (45) 172 (49)

C19(rectosigmoidjunction)

393(24)

112 (25) 124 (23) 81 (23)

C20 (rectum) 406(25)

86 (19) 163 (30) 93 (27)

C21 (anus & analcanal)

50 (3) 6 (1) 14 (3) 4 (1)

Male 834(51)

213 (48) 314 (58) 193 (55)

Female 816(49)

233 (52) 230 (42) 157 (45)

Age

<50 122(7)

2 (0.5) 1 (0.2) 3 (1.)

50–59 238(14)

15 (3) 16 (3) 44 (13)

60–69 452(27)

66 (15) 137 (25) 101 (29)

70–79 481(29)

167 (37) 238 (44) 134 (38)

80+ 357(22)

196 (44) 152 (28) 68 (19)

SIMD quintiles of socio-economic deprivation

1 (poorest) 215(13)

66 (15) 75 (14) 42 (12)

2 249(15)

76 (17) 83 (16) 56 (16)

3 285(17)

78 (18) 81 (15) 55 (16)

4 554(34)

144 (33) 186 (35) 132 (38)

5 (most affluent) 327(20)

78 (18) 110 (21) 60 (17)

Duke’s stage

A 174(11)

36 (8) 66 (12) 71 (20)

B 382(23)

83 (19) 166 (31) 129 (37)

C 470(28)

90 (20) 146 (27) 107 (31)

D 323(20)

94 (21) 51 (9) 10 (3)

Unknown 301(18)

143 (32) 115 (21) 33 (9)

All-causemortality

1101(67)

391 (88) 353 (65) 153 (49)

Colorectalmortality

601(36)

193 (43) 171 (31) 56 (16)

Table 1Characteristics of the cohort.

n (%) All-causemortality(%)

Colorectalmortality(%)

All Patients 2990 1998 (67) 1021 (34)

ICD10 classification

C18 (colon) 1458(49)

979 (67) 475 (33)

C19(rectosigmoidjunction)

710 (24) 506 (71) 267 (38)

C20 (rectum) 748 (25) 470 (63) 253 (34)C21 (anus &

anal canal)74 (2) 42 (57) 25 (34)

Male 1554(52)

1019 (66) 529 (34)

Female 1436(48)

979 (68) 492 (34)

Age

<50 128 (4) 58 (45) 148 (37)50–59 313 (10) 137 (44) 154 (33)60–69 756 (25) 414 (55) 169 (34)70–79 1020

(34)719 (70) 359 (35)

80+ 773 (26) 670 (87) 177 (31)

Scottish index of multiple deprivation (SIMD) quintiles of socio-

economic deprivation

1 (poorest) 398 (14) 285 (72) 148 (37)2 464 (16) 344 (74) 154 (33)3 499 (17) 334 (66) 169 (34)4 1016

(34)652 (64) 359 (35)

5 (most affluent) 575 (19) 353 (61) 177 (31)

Dukes stage

A 347 (12) 125 (36) 26 (7)B 760 (25) 359 (47) 127 (17)C 813 (27) 546 (67) 299 (37)D 478 (16) 465 (97) 388 (66)Unknown 592 (20) 503 (85) 254 (43)

1052 C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057

with increased risk, whilst the top two quintiles forsocial deprivation were at reduced risk compared tothe most deprived group.

Aspirin use pre-diagnosis suggested an increased riskof mortality but this effect was lost after adjustment forother factors. To determine whether aspirin use pre-diagnosis was associated with mortality a model withan interactive term to test for an effect of pre-diagnosisaspirin on post-diagnosis use was created but showedno effect existed (HR = 1.02, 95% CI 0.70–1.47,p = 0.934).

To examine the effect of adherence to aspirin on all-cause mortality patients were classed as having highadherence for values of 80% or above (715 patients,80% of patients taking aspirin post-diagnosis) and lowadherence otherwise (179 patients, 20%) and comparedto those patients not taking aspirin. After adjusting forgender, age, social class, Dukes’ stage and aspirin usepre-diagnosis patients with low adherence had a reduced

hazard ratio for death (HR = 0.37, 95% CI 0.30–0.46,p < 0.001) as did for those with high adherence(HR = 0.47, 95% CI 0.42–0.54, p < 0.001). Examining

Table 3Cox proportional hazard model all cause mortality.

Unadjusted hazardratio (HR) (95%confidence interval(CI)), p-value

Adjusted HR (95% CI) ,p-value

Female 1.0 1.0

Male 0.92 (0.84–1.00), 0.056 1.03 (0.94–1.13), 0.530Age (each

additionalyear)

1.04 (1.04–1.04),<0.001 1.04 (1.03–1.04),<0.001

SIMD quintiles of socio-economic deprivation

1 (poorest) 1.0 1.02 1.02 (0.88–1.20), 0.764 0.94 (0.81–1.10), 0.4623 0.88 (0.74–1.02), 0.087 0.86 (0.73–1.01), 0.0614 0.79 (0.68–0.90), 0.001 0.79 (0.69–0.91), 0.0015 (most

affluent)0.73 (0.62–0.85),<0.001 0.73 (0.63–0.86),<0.001

Dukes’ stage

A 1.0 1.0B 1.43(1.17–1.75), 0.001 1.31(1.07–1.61), 0.010C 2.43 (2.00–2.95),<0.001 2.24 (1.84–2.73),<0.001D 9.00 (7.36–

11.00),<0.0018.42 (6.88–10.32),<0.001

Unknown 5.22 (4.29–6.35),<0.001 4.08 (3.34–4.99),<0.001

No aspirinpre-diagnosis

1.0 1.0

Aspirinpre-diagnosis

1.30 (1.19–1.43),<0.001 0.99 (0.90–1.09), 0.861

No aspirinpost-diagnosis

1.0 1.0

On aspirin 0.64 (0.55–0.75),<0.001 0.67 (0.57–0.79),<0.001

Table 4Cox proportional hazard model colorectal mortality.

Unadjusted hazardratio (HR) (95%confidence interval(CI)), p-value

Adjusted HR (95% CI) ,p-value

Female 1.0 1.0Male 0.95 (0.84–1.07), 0.411 1.00 (0.88–1.13), 0.944

Age (eachadditionalyear)

1.03 (1.02–1.03),<0.001 1.03 (1.02–1.03),<0.001

SIMD quintiles of socio-economic deprivation

1 (poorest) 1.0 1.02 0.89 (0.71–1.12), 0.323 0.83 (0.66–1.04), 0.0993 0.88 (0.69–1.07), 0.197 0.87 (0.70–1.09), 0.2244 0.80 (0.69–1.02), 0.012 0.84 (0.70–1.02), 0.0815 (most

affluent)0.74 (0.57–0.89), 0.003 0.75 (0.60–0.93), 0.009

Dukes’ stage

A 1.0 1.0B 2.43 (1.60–3.71),<0.001 2.25 (1.48–3.44),<0.001C 6.40 (4.29–9.56),<0.001 5.95 (3.98–8.89),<0.001D 30.38 (20.30–

45.47),<0.00128.77 (19.21–43.09),<0.001

Unknown 12.85 (8.57–19.25),<0.001

10.72 (7.13–16.10),<0.001

No aspirinpre-diagnosis

1.0 1.0

Aspirinpre-diagnosis

1.20 (1.06–1.36), 0.005 0.96 (0.84–1.11), 0.596

No aspirin 1.0 1.0On aspirin 0.50 (0.39–0.64),<0.001 0.58 (0.45–0.75),<0.001

C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057 1053

only those patients who used aspirin post-diagnosisshowed no difference for high against low adherenceon risk of death (HR = 1.23, 95% CI 0.98–1.54,p < 0.070).

Summing all periods of aspirin use post-diagnosisand entering this as the number of years of treatmentinto a model for all-cause mortality demonstrated areduced hazard (HR = 0.67, 95% CI 0.63–0.71,p < 0.001) after adjustment for gender, age, social class,Dukes’ stage and aspirin use pre-diagnosis.

Including the site of individual cancers into the modelshowed no difference in the reduced hazard for aspirinuse post-diagnosis (HR = 0.67, 95% CI 0.57–0.79,p < 0.001), although patients with cancers of the rectum(ICD-10 C20) (HR = 0.88, 95% CI 0.79–0.99, p = 0.026)or anus and anal canal (ICD-10 C21) (HR = 0.59,95% CI 0.43–0.82, p = 0.001) had reduced mortalityhazard compared to colon patients (ICD10- C18). How-ever, there was no difference in mortality hazard forpatients with cancer of the rectosigmoid junction(ICD-10 C19) (HR = 0.99, 95% CI 0.88–1.10,p = 0.747).

Running separate models for each cancer site showeda significant reduction in hazard for aspirin use post-diagnosis after adjustment for gender, age, social class,

Dukes’ stage and aspirin use pre-diagnosis for patientswith cancer of the colon (ICD-10 C18) (HR = 0.72,95% CI 0.57–0.91, p = 0.006) and rectosigmoid junction(ICD-10 C19) (HR = 0.51, 95% CI 0.36–0.74, p < 0.001)but not for cancers of the rectum (ICD-10 C20)(HR = 0.80, 95% CI 0.58–1.11, p = 0.186) or anus andanal canal (ICD-10 C21) (HR = 0.66, 95% CI 0.17–2.57, p = 0.549).

Cox proportional hazard models were also createdwith colorectal cancer deaths as the outcome (seeTable 4). After adjustment for gender, age, social class,Dukes’ stage and aspirin use pre-diagnosis aspirin usepost-diagnosis was associated with lower risk of deathfrom colorectal cancer (HR = 0.58, 95% CI 0.45–0.75,p < 0.001). Increasing age and stage at diagnosis wereassociated with increased risk of death from colorectalcancer, whilst the most affluent quintiles of patients wereat reduced risk compared to the most deprived group.

6. Discussion

Our data suggest that taking aspirin post-diagnosismay reduce the risk of all-cause and colorectal cancer

1054 C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057

specific mortality. This effect is independent of age, siteand stage of disease and social deprivation which areall also associated with survival. Adherence to aspirinpost-diagnosis of colorectal cancer is high. Approxi-mately one in three patients with colorectal cancer aretaking aspirin post-diagnosis and each additional yearof aspirin therapy decreases their risk of mortality com-pared to non-users.

6.1. Comparison to literature

This study confirms the findings in the literature thatpost-diagnosis aspirin use increases all-cause and colo-rectal cancer specific survival. The magnitude of effectthat we saw for all-cause mortality (adjusted HR 0.67,95% CI 0.57–0.79) was more marked than that seen inthe study by Bastiaannet et al. (HR 0.77, 95% CI 0.63–0.95) and for colorectal cancer mortality (adjusted HR0.58, 95% CI 0.45–0.75) than Chan et al. (HR 0.79,95% CI 0.53–0.95).28,29 Looking at the site of diseasewe showed a significant survival benefit for aspirin usepost-diagnosis for patients with cancer of the colonand rectosigmoid junction. There was no such signifi-cant difference in cancers of the rectum or anus and analcanal although adjusting for site within our main analy-sis showed no difference to the effect of post-diagnosisaspirin. This is in contrast to Bastiaannet et al. whoreported no benefit in patients with cancers of the rec-tum.29 However, it was uncertain how these cancerswere classified and their reporting of exposure to aspirinwas based on frequency of prescription refills whereaswe used a more robust method of comparing periodsof exposure to non-exposure. Differences in the effectof aspirin based on cancer site require further analysisto determine if there is a positive association withimproved survival.

In contrast to the use of aspirin after diagnosis, its usebefore diagnosis was not associated with a significantchange in either colorectal cancer-specific mortality oroverall mortality. When patients who were taking aspirinboth before and after diagnosis were excluded from theanalysis the effect of aspirin on disease specific mortalityappeared to be enhanced. However, a formal test forinteraction between aspirin use before diagnosis andaspirin use after diagnosis was not statistically signifi-cant. In contrast, in an analysis of women aged underthe age of 85 years from the California Teachers Studypre-diagnosis use of nonsteroidal anti-inflammatorydrug (NSAID) for greater than five years in those devel-oping colorectal cancer was associated with improvedoverall and disease specific survival compared with noaspirin use pre-diagnosis.31 Furthermore, in anotherstudy utilising the population-based Puget Sound sur-veillance, epidemiology and end results (SEER) registryin the United States, NSAID use prior to a diagnosisof colorectal cancer was associated with around a 20%lower rate of disease-specific mortality.32 Patients on

regular aspirin who develop adenocarcinomas are lesslikely to have metastatic disease at presentation and areless likely to develop metastatic disease subsequently,however there is no clear evidence of benefit from pre-diagnostic aspirin use in patients with colorectalcancer.33,34

The findings in our study were very much in accordwith those from Chan et al. and Bastiaannet et al. aswe could demonstrate no effect of aspirin pre-diagnosison either all cause or colorectal cancer specific mortalityand we could find no evidence of an interaction betweenpre-diagnosis and post-diagnosis of aspirin. From thestandpoint of using aspirin as a therapeutic agent incolorectal cancer it is reassuring to note that our studydemonstrated not only an improved cancer specific sur-vival but also an improved overall survival; an impor-tant observation given the potential hazards of chronicaspirin intake.35

The level of aspirin use in our population was 45%which is slightly higher compared to the 36% estimatefor an American general population.36 This may beexplained by the increased age of our study populationor geographical differences in use of prophylactic aspirinto reduce cardiovascular disease.

6.2. Strengths and limitations

The study looked at all colorectal cancers diagnosedfrom an entire population. The study used dispensedprescribing records from a closed prescribing system,where every prescription presented to a pharmacist issubsequently recorded and collated, rather than patientrecall to determine aspirin usage. These data were thenlinked to official cancer registry records and health boardpopulation databases to provide the study dataset.

We assessed adherence to aspirin and found it wasvery high with 80% of patients taking tablets 4 daysout of 5. However using less than this as a marker oflow adherence patients still had a reduced risk of death.This reduced risk was greater than those patients withhigh adherence although there was no statistical differ-ence between the two groups.

Over the counter prescribing of aspirin was unac-counted for in the study but previous work in Taysideestimated over 94% of 75 mg aspirin tablets used in1998 were from filled prescriptions.37 However, we donot know why aspirin was indicated for individualpatients. We also could not tell whether the individualprescribed the aspirin subsequently took it, they mayhave stored it or given it to other family members. How-ever, as the majority of our aspirin was a low 75 mg dosewe do not think it would have been used for pain relieffor family members and the high adherence we notedwould also suggest patients were taking the medication.

There are a number of other factors for which wewere unable to adjust – comorbidity, smoking habits,BMI etc. Use of aspirin pre-diagnosis is likely to be

C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057 1055

the marker of existing cardiovascular disease, whichmay explain why it initially seemed to be associated withpoorer prognosis. We did not have findings from post-diagnostic colonoscopy within our dataset but standardpractice is for all patients to have a colonoscopy at2 years after diagnosis and then 2 years later if polypsare present, otherwise the second post-treatment colon-oscopy is 7 years after diagnosis. As we were able tolook at individual level data for a complete populationunless there were differences in each risk factor betweenpatients using aspirin and those not it is unlikely theeffect shown is caused solely by unknown confounders,although the size of the effect may be different.

Aspirin use may be seen as a proxy for patientsadopting a healthier lifestyle but we feel this is unlikelyto explain completely the association between aspirinuse and reduced risk of mortality. The increased effectof aspirin on reducing mortality due to colorectal cancerwould support the theory that the aspirin acts on thecancer at a basic molecular level.

We did not explicitly explore potential harms causedby aspirin – haemorrhage and digestive disorders - butdid examine the effect of co-prescribing of proton pumpinhibitors with aspirin post-diagnosis. We still found apositive effect of aspirin on all-cause mortality (HR foraspirin use after allowing for PPI use = 0.68, 95% CI0.57–0.80, p < 0.001).

6.3. Clinical implications

The mechanism whereby aspirin may exert a thera-peutic effect is important, particularly if analysis ofPTGS expression might indicate those who wouldbenefit. Aspirin has complex effects upon prostaglandinbiosynthesis and many of these effects may be dose-dependent. Findings from experiments using high dosesof aspirin may not be applicable to clinical studiesemploying lower (75 mg) doses of aspirin. Chuladaet al. showed that both PTGS-1 and PTGS-2 wereinvolved in polyp formation in Min/+ mice and thisimplies that disrupting either or both isoforms of cyclo-oxygenase (COX-1 and COX-2) might protect againstcarcinogenesis.25 Confirmation of this idea comes fromthe recent finding that polymorphisms in the PTGS-1gene may influence survival in patients with colorectalcancer.26 The recently reported VICTOR trial of theselective PTGS 2 inhibitor rofecoxib in stage 2 and 3colorectal cancer showed no effect on either overall sur-vival or recurrence.38 There was a significant trendtowards prevention of recurrence by rofecoxib duringthe first year. This trial was terminated early becauseof worldwide withdrawal of rofecoxib (median durationof treatment 7.4 months) but the findings are consistentwith the hypothesis that PTGS 2 inhibition delays orprevents progression of colorectal cancer.

Rofecoxib was withdrawn because of an increasedincidence of cardiovascular events39 but aspirin does

not have this drawback and although long-term lowdose aspirin is associated with an increased risk of pepticulcer and gastrointestinal bleeding40 (which may beameliorated by the use of proton pump inhibitors41)its safety profile is such that it widely used for the pre-vention of cardiovascular events.42 In addition, there isevidence from a meta-analysis of polyp prevention trialsthat aspirin did not increase the risk of major bleeding,although it did increase the risk of haemorrhagicstroke.5

Clearly it would be of great interest to conduct pro-spective studies to assess the efficacy of aspirin and otherNSAIDs as adjuvant treatment in colorectal cancer. Theideal dosage for such a study is not clear; currently theaspirin for Dukes C and high risk Dukes B colorectalcancers (ASCOLT) study, based in Singapore, is rando-mising patients with Duke’s Stage C and high risk B toplacebo or 200 mg of aspirin per day for 3 years aftercompletion of standard adjuvant therapy.43 However,there is now good evidence that aspirin at 75 mg is effec-tive in preventing colorectal adenomas and indeed onetrial comparing 81 and 325 mg showed an effect onlywith the lower dose.1 In addition the study by Rothwellet al. on RCTs of aspirin use in the prevention of vascu-lar events has indicated that doses of aspirin greaterthan 75 mg per day may be effective in preventing colo-rectal cancer11 and this has been confirmed by a casecontrol study of 2279 patients with colorectal cancer.44

The majority of aspirin used in this study and the onereported by Bastiaannet et al. was at 75 and 80 mg dos-age.29 A trial in the US/Canada is also examining theuse of 400 mg daily of celecoxib against placebo inpatients with Duke’s Stage C.45

7. Conclusions

Our study suggests that aspirin use post-diagnosis ofcolorectal cancer may reduce both all cause andcolorectal cancer specific mortality and that a low dose(75 mg) of the drug is effective. However further work,through randomised trials designed to test the effective-ness of low dose aspirin post-diagnosis in stage 2 and 3colorectal cancer, is needed both in those receiving con-ventional adjuvant chemotherapy and in those forwhom, because of age or comorbidity, it might beinappropriate.

Contributors

R.S. conceived the study and planned it with C.McC.,A.M. and P.D. C.McC carried out the analysis with sup-port from P.D. All authors contributed to the writing ofthe paper. C.McC is the guarantor.

Funding

There was no external funding for this work.

1056 C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057

Ethical approval

Not required; all data were fully anonymised anddata use was compliant with the Health InformaticsCentre research governance process which are approvedby the Tayside Committee on Medical Research Ethicsand the Caldicott Guardian.

Conflict of interest statement

All authors have completed the ICMJE uniform dis-closure form at www.icmje.org/coi_disclosure.pdf; noneof the authors have financial relationships with anyorganisations that might have an interest in the submit-ted work in the previous three years and have no otherrelationships or activities that could appear to haveinfluenced the submitted work.

Acknowledgements

We thank Alison Bell, Chris Hall and DuncanHeather of the Health Informatics Centre, Universityof Dundee, for anonymisation, record linkage, andother procedural assistance in preparing the dataset.

References

1. Baron JA, Cole BF, Sandler RS, et al. A randomized trial ofaspirin to prevent colorectal adenomas. Obstet Gynecol Surv

2003;58(8):538–9.2. Sandler RS, Halabi S, Baron JA, et al. A randomized trial of

aspirin to prevent colorectal adenomas in patients with previouscolorectal cancer. Obstet Gynecol Surv 2003;58(8):539–41.

3. Benamouzig R, Deyra J, Martin A, et al. Daily soluble aspirin andprevention of colorectal adenoma recurrence: one-year results ofthe APACC trial. Gastroenterology 2003;125(2):328–36.

4. Logan RFA, Grainge MJ, Shepherd VC, Armitage NC, Muir KR.Aspirin and folic acid for the prevention of recurrent colorectaladenomas. Gastroenterology 2008;134(1):29–38.

5. Cole BF, Logan RF, Halabi S, et al. Aspirin for the chemopre-vention of colorectal adenomas: meta-analysis of the randomizedtrials. J Natl Cancer Inst 2009;101(4):256–66.

6. Bertagnolli MM, Eagle CJ, Zauber AG, et al. Celecoxib for theprevention of sporadic colorectal adenomas. N Engl J Med

2006;355(9):873–84.7. Arber N, Eagle CJ, Spicak J, et al. Celecoxib for the prevention of

colorectal adenomatous polyps. N Engl J Med 2006;355(9):885–95.8. Baron JA, Sandler RS, Bresalier RS, et al. A randomized trial of

rofecoxib for the chemoprevention of colorectal adenomas.Gastroenterology 2006;131(6):1674–82.

9. Leslie A, Carey FA, Pratt NR, Steele RJC. The colorectaladenoma–carcinoma sequence. Br J Surg 2002;89(7):845–60.

10. Flossmann E, Rothwell PM; British Doctors Aspirin Trial and theUK-TIA Aspirin Trial. Effect of aspirin on long-term risk ofcolorectal cancer: consistent evidence from randomised andobservational studies. Lancet 2007;369(9573):1603–13.

11. Rothwell PM, Fowkes FG, Belch JF, et al. Effect of daily aspirinon long-term risk of death due to cancer: analysis of individualpatient data from randomised trials. Lancet 2011;377(9759):31–41.

12. Burn J, Gerdes A-M, Macrae F, et al. Long-term effect of aspirinon cancer risk in carriers of hereditary colorectal cancer: ananalysis from the CAPP2 randomised controlled trial. Lancet

2011;378(9809):2081–7.

13. Chan AT, Ogino S, Fuchs CS. Aspirin and the risk of colorectalcancer in relation to the expression of COX-2. N Engl J Med

2007;356(21):2131–42.14. Markowitz SD. Aspirin and colon cancer – targeting prevention?

N Engl J Med 2007;356(21):2195–8.15. Eberhart CE, Coffey RJ, Radhika A, et al. Up-regulation of

cyclooxygenase 2 gene expression in human colorectal adenomasand adenocarcinomas. Gastroenterology 1994;107(4):1183–8.

16. Soumaoro LT, Uetake H, Higuchi T, et al. Cyclooxygenase-2expression. Clin Cancer Res 2004;10(24):8465–71.

17. Brown JR, DuBois RN. COX-2: a molecular target for colorectalcancer prevention. J Clin Oncol 2005;23(12):2840–55.

18. Tomozawa S, Tsuno NH, Sunami E, et al. Cyclooxygenase-2overexpression correlates with tumour recurrence, especiallyhaematogenous metastasis, of colorectal cancer. Br J Cancer

2000;83(3):324–8.19. Williams CS, Watson AJM, Sheng H, et al. Celecoxib prevents

tumor growth in vivo without toxicity to normal gut: lack ofcorrelation between in vitro and in vivo models. Cancer Res

2000;60(21):6045–51.20. Williams CS, Sheng H, Brockman JA, et al. A cyclooxygenase-2

inhibitor (SC-58125) blocks growth of established human coloncancer xenografts. Neoplasia 2001;3(5):428–36.

21. Yamauchi T, Watanabe M, Hasegawa H, et al. The potential for aselective cyclooxygenase-2 inhibitor in the prevention of livermetastasis in human colorectal cancer. Anticancer Res

2003;23(1A):245–9.22. Yao M, Kargman S, Lam EC, et al. Inhibition of cyclooxygenase-

2 by rofecoxib attenuates the growth and metastatic potential ofcolorectal carcinoma in mice. Cancer Res 2003;63(3):586–92.

23. Yao M, Lam EC, Kelly CR, Zhou W, Wolfe MM. Cyclooxygen-ase-2 selective inhibition with NS-398 suppresses proliferation andinvasiveness and delays liver metastasis in colorectal cancer. Br J

Cancer 2004;90(3):712–9.24. Yao M, Zhou W, Sangha S, et al. Effects of nonselective

cyclooxygenase inhibition with low-dose ibuprofen on tumorgrowth, angiogenesis, metastasis, and survival in a mouse model ofcolorectal cancer. Clin Cancer Res 2005;11(4):1618–28.

25. Chulada PC, Thompson MB, Mahler JF, et al. Genetic disruptionof Ptgs-1, as well as of Ptgs-2, reduces intestinal tumorigenesis inmin mice. Cancer Res 2000;60(17):4705–8.

26. Coghill AE, Newcomb PA, Poole EM, et al. Genetic variation ininflammatory pathways is related to colorectal cancer survival.Clin Cancer Res 2011;17(22):7139–47.

27. Fuchs C, Meyerhardt J, Heseltine D, et al. Influence of regularaspirin use on survival of patients with stage III colon cancer:findings from intergroup trial CALGB 89803. J Clin Oncol

2005;23(Suppl. 16):3530.28. Chan AT, Ogino S, Fuchs CS. Aspirin use and survival after

diagnosis of colorectal cancer. JAMA 2009;302(6):649–58.29. Bastiaannet E, Sampieri K, Dekkers OM, et al. Use of aspirin

postdiagnosis improves survival for colon cancer patients. Br J

Cancer 2012;106(9):1564–70.30. McCowan C, Shearer J, Donnan PT, et al. Cohort study

examining tamoxifen adherence and its relationship to mortalityin women with breast cancer. Br J Cancer 2008;99(11):1763–8.

31. Zell JA, Ziogas A, Bernstein L, et al. Nonsteroidal anti-inflam-matory drugs. Cancer 2009;115(24):5662–71.

32. Coghill AE, Newcomb PA, Campbell PT, et al. Prediagnostic non-steroidal anti-inflammatory drug use and survival after diagnosisof colorectal cancer. Gut 2011;60(4):491–8.

33. Rothwell PM, Wilson M, Price JF, et al. Effect of daily aspirin onrisk of cancer metastasis: a study of incident cancers duringrandomised controlled trials. Lancet 2012;379(9826):1591–601.

34. Algra AM, Rothwell PM. Effects of regular aspirin on long-termcancer incidence and metastasis: a systematic comparison ofevidence from observational studies versus randomised trials.Lancet Oncol 2012;13(5):518–27.

C. McCowan et al. / European Journal of Cancer 49 (2013) 1049–1057 1057

35. Dube C, Rostom A, Lewin G, et al. The use of aspirin for primaryprevention of colorectal cancer: a systematic review prepared forthe U.S. preventive services task force. Ann Intern Med

2007;146(5):365–75.36. Ajani UA, Ford ES, Greenland KJ, Giles WH, Mokdad AH.

Aspirin use among U.S. adults: behavioral risk factor surveillancesystem. Am J Prev Med 2006;30(1):74–7.

37. Morant SV, McMahon AD, Cleland JGF, Davey PG, MacDon-ald TM. Cardiovascular prophylaxis with aspirin: costs of supplyand management of upper gastrointestinal and renal toxicity. Br J

Clin Pharmacol 2004;57(2):188–98.38. Midgley RS, McConkey CC, Johnstone EC, et al. Phase III

randomized trial assessing rofecoxib in the adjuvant setting ofcolorectal cancer: final results of the VICTOR trial. J Clin Oncol

2010;28(30):4575–80.39. Kerr DJ, Dunn JA, Langman MJ, et al. Rofecoxib and cardio-

vascular adverse events in adjuvant treatment of colorectal cancer.N Engl J Med 2007;357(4):360–9.

40. Yeomans ND. Aspirin: old drug, new uses and challenges. J

Gastroenterol Hepatol 2011;26(3):426–31.41. Lai KC, Lam SK, Chu KM, et al. Lansoprazole for the prevention

of recurrences of ulcer complications from long-term low-doseaspirin use. N Engl J Med 2002;346(26):2033–8.

42. Hennekens CH, Dyken ML, Fuster V. Aspirin as a therapeuticagent in cardiovascular disease: a statement for healthcareprofessionals from the American Heart Association. Circulation

1997;96(8):2751–3.43. INDOX Cancer Research Network. ASCOLT study. <http://

indox.org.uk/research/clinical/ascolt>; 2010.44. Din FVN, Theodoratou E, Farrington SM, et al. Effect of aspirin

and NSAIDs on risk and survival from colorectal cancer. Gut

2010;59(12):1670–9.45. National Cancer Institute (NCI). Oxaliplatin, leucovorin calcium,

and fluorouracil with or without celecoxib in treating patients with

stage III colon cancer previously treated with surgery. <http://clinicaltrials.gov/ct2/show/NCT01150045>; 2012.