2013 cancer de pancreas y aas articulo original

A Prospective Study of Aspirin Use and the Risk ofPancreatic Cancer in Women

Eva S. Schernhammer, Jae-Hee Kang, Andrew T. Chan, Dominique S.Michaud, Halcyon G. Skinner, Edward Giovannucci, Graham A. Colditz,Charles S. Fuchs

Background: In vitro experiments and limited animal studiessuggest that aspirin and nonsteroidal anti-inflammatorydrugs may inhibit pancreatic carcinogenesis. Because fewstudies have examined the association between aspirin useand pancreatic cancer in humans and the results have beeninconsistent, we examined the relationship between aspirinuse and the development of pancreatic cancer in the Nurses’Health Study. Methods: Among 88 378 women without can-cer at baseline, we documented 161 cases of pancreatic can-cer during 18 years of follow-up. Aspirin use was first as-sessed at baseline in 1980 and updated biennially thereafter.All statistical tests were two-sided. Results: Participants wereclassified according to history of aspirin use. In a multivari-able analysis, the risk of pancreatic cancer was not associ-ated with current regular aspirin use (defined as two or morestandard tablets per week; relative risk [RR] � 1.20, 95%confidence interval [CI] � 0.87 to 1.65), compared with useof fewer than two tablets per week. Increasing duration ofregular aspirin use, compared with non-use, was associatedwith a statistically significant increase in risk: Women whoreported more than 20 years of regular aspirin use had anincreased risk of pancreatic cancer (RR � 1.58, 95% CI �1.03 to 2.43; Ptrend � .01). Among women who reportedaspirin use on at least two of three consecutive biennialquestionnaires compared with consistent non-users of aspi-rin, the risk increased with dose (one to three tablets perweek: RR � 1.11, 95% CI � 0.70 to 1.76; four to six tabletsper week: RR � 1.29, 95% CI � 0.70 to 2.40; seven to 13tablets per week: RR � 1.41, 95% CI � 0.76 to 2.61; and>14 tablets per week: RR � 1.86, 95% CI � 1.03 to 3.35)(Ptrend � .02). Conclusion: Extended periods of regular as-pirin use appear to be associated with a statistically signifi-cantly increased risk of pancreatic cancer among women.[J Natl Cancer Inst 2004;96:22–8]

Pancreatic cancer, the fourth leading cause of cancer-relatedmortality in the United States (1), is a rapidly fatal malignancywith limited effective treatment. Nonetheless, other than ciga-rette smoking, few risk factors have been consistently linked tothe risk of pancreatic cancer. Clearly, further studies are neededto identify potential causes and chemopreventive agents for thedisease.

There is considerable evidence that nonsteroidal anti-inflammatory drugs (NSAIDs), particularly aspirin, reduce therisk of several cancers and premalignant lesions. Observationaland intervention studies (2) consistently demonstrate the valueof these drugs as chemopreventive agents for colorectal neopla-sia. Although the mechanism by which aspirin reduces the riskof neoplasms remains unclear, the oncostatic effect is thought tobe based in part on its anti-inflammatory actions, mediated by

the inhibition of cyclooxygenase 2: NSAIDs indirectly inhibitthe biosynthesis of prostaglandins by blocking the enzyme pros-taglandin G/H synthetase (3). Two isoforms of cyclooxygenasehave been identified: cyclooxygenase 1 and cyclooxygenase 2(4,5). Although cyclooxygenase 1 plays a central role in plateletaggregation and protection of stomach mucosa and is expressedin most tissues, cyclooxygenase 2 is induced by various cyto-kines and mitogenic factors and is found in high concentrationsin tissues undergoing inflammatory and wound healing pro-cesses as well as in some neoplasms (6). Both cyclooxygenase 1and cyclooxygenase 2 initiate the production of biologicallyimportant prostanoids, which are involved in cell signaling (7).Numerous experimental studies (3) also indicate that NSAIDscan influence tumorigenesis, apoptosis, and angiogenesis.

In vitro experiments and limited animal studies (8,9) suggestthat aspirin and NSAIDs inhibit pancreatic carcinogenesis. Todate, few studies have examined the associations between anal-gesic use and pancreatic cancer in humans, and the results havebeen inconsistent. A recent prospective study of U.S. women(10) reported a statistically significant inverse association be-tween current aspirin use and the risk of pancreatic cancer,although the analysis was limited by the small number of pa-tients studied (n � 80). Moreover, updated aspirin use (beyondthe baseline measure) and past use were not assessed.

We conducted a prospective study in a large cohort of womenwith 18 years of follow-up and detailed periodic assessment ofaspirin use to further define the possible preventive effects ofaspirin for pancreatic cancer with particular emphasis on theeffects of dosage and duration of aspirin use. In this cohort,aspirin data were elicited before any diagnosis of pancreaticcancer, thus avoiding potential biases that may occur whenobtaining such information from pancreatic cancer patients ortheir next of kin.

Affiliations of authors: Channing Laboratory, Department of Medicine,Brigham and Women’s Hospital and Harvard Medical School, Boston, MA(ESS, JHK, ATC, HGS, EG, GAC CSF); Ludwig Boltzmann-Institute forApplied Cancer Research, KFJ-Spital, Vienna, Austria (ESS); GastrointestinalUnit, Massachusetts General Hospital, Boston (ATC); Nutritional EpidemiologyBranch, National Cancer Institute, Rockville, MD (DSM); Departments ofNutrition (EG) and Epidemiology (HGS, GAC), Harvard School of PublicHealth, Boston; Harvard Center for Cancer Prevention, Boston (GAC); Epide-miology Program (GAC) and Department of Adult Oncology (CSF), Dana-Farber Cancer Institute, Boston.

Correspondence to: Eva S. Schernhammer, MD, DrPH, Channing Laboratory,181 Longwood Ave., Boston, MA 02115 (e-mail: [email protected]).

See “Notes” following “References.”

DOI: 10.1093/jnci/djh001Journal of the National Cancer Institute, Vol. 96, No. 1, © Oxford UniversityPress 2004, all rights reserved.

22 ARTICLES Journal of the National Cancer Institute, Vol. 96, No. 1, January 7, 2004

by guest on April 27, 2013

http://jnci.oxfordjournals.org/D

ownloaded from

http://jnci.oxfordjournals.org/

PATIENTS AND METHODS

Study Cohort

In 1976, the Nurses’ Health Study enrolled 121 700 femaleregistered nurses aged 30 to 55 years to gather, through mailedquestionnaires, information on their health status, medical his-tory, and known and suspected risk factors for cancer andcoronary heart disease. Follow-up questionnaires were sent bi-ennially to cohort participants thereafter. Further details of thecohort are reported elsewhere (11–13).

Ascertainment of Aspirin and NSAID Use

Information on aspirin use was first collected at baseline in1980 and updated biennially thereafter with the exception of1986. In 1980, we ascertained current aspirin use (defined asaspirin use in most weeks) from the study participants andprovided space for participants to fill in the number of tabletstaken each week and to indicate for how many years they hadused aspirin. In 1982, we asked if they currently took aspirin atleast once per week and provided prespecified categories of thetotal amount of aspirin use per week (1–3, 4–6, 7–14, or �15adult tablets). In 1984 and 1988, we queried the average numberof days per month of aspirin use (none, 1–4, 5–14, 15–21, or�22 days/month) and, on days they did take aspirin, the numberof tablets usually taken (never take it, 1, 2, 3–4, 5–6, or �7tablets). In 1990 and 1992, only the number of days per monthof aspirin use was assessed (none, 1–4, 5–14, 15–21, or �22days/month) but not the number of aspirin tablets per day. In1994 and 1996, women were asked whether they had regularlyused aspirin (two or more times per week) in the past 2 years,how frequently they had used aspirin (�1 day/month, 1–3 days/month, 1–2 days/week, 3–4 days/week, 5–6 days/week, ordaily) and how many tablets per week, on average (0, 0.5–2,3–5, 6–14, or �15 tablets per week). Early in the study, mostwomen used standard-dose tablets (325 mg per tablet); however,to reflect overall trends in consumption of low-dose or babyaspirin (81 mg per tablet), the 1994 and 1996 questionnairesasked participants to convert use of four baby aspirin to oneadult tablet when choosing their responses. Some regrouping ofresponses, therefore, was required to adjust for the differingways in which aspirin-use habits were recorded between 1980and 1998 (14). On the basis of a previous analysis of this cohort,women who reported taking two or more standard aspirin tabletsper week were defined as regular users, whereas those whoreported less aspirin use were defined as non-regular users (14).To elicit the reasons for aspirin use, a short questionnaire wassent in 1990 to 100 participants in the Nurses’ Health Study whoreported taking one to six aspirin per week (90% response) andto 100 women who reported taking seven or more aspirin perweek (92% response) on the 1980, 1982, and 1984 question-naires (14). The major reasons for use among women taking oneto six aspirin and seven or more aspirin per week were headache(32% and 18%, respectively), arthritis and other musculoskeletalpain (30% and 50%), a combination of headache and musculo-skeletal pain (16% and 15%), cardiovascular disease prevention(9% and 8%), and other reasons (13% and 9%) (14).

In 1980, participants were also asked whether they werecurrently taking nonsteroidal analgesics other than aspirin(Motrin, Indocin, Tolectin, and Clinoril) in most weeks. Inaddition, users were asked to provide the number of years they

had used the drugs and the number of tablets taken per week.Non-aspirin NSAID use was not reassessed until 1990. In 1990and 1992, we asked for the frequency of non-aspirin NSAID use(none, 1–4, 5–14, 15–21, or �22 days per month) and, in 1994and 1996, we assessed regular use of non-aspirin NSAIDs (twoor more times per week).

Smoking History and Other Risk Factors

Smoking status and history of smoking were obtained atbaseline and in all subsequent questionnaires. Current smokersalso reported intensity of smoking (average number of cigarettessmoked per day) on each questionnaire. Past smokers reportedwhen they last smoked, and the time since quitting was calcu-lated for those who quit during follow-up. In a previous study(15), we examined the relationship between smoking and pan-creatic cancer risk in detail; the strongest associations wereobserved in analyses of total pack-years smoked within theprevious 15 years. Participants were asked about history ofdiabetes at baseline and in all subsequent questionnaires. Be-cause pancreatic cancer is frequently associated with profoundweight loss, we did not adjust for weight change or most recentbody mass index but used baseline body mass index (1976, thestart of the cohort), which was the most important weight riskfactor for pancreatic cancer in this cohort (16). Other questionsrelevant to the association between aspirin use and pancreaticcancer provided information on the study participants’ age andphysical activity.

Ascertainment of Pancreatic Cancer and Deaths

We included pancreatic cancers reported on the biennialquestionnaires between the return of the 1980 questionnaire andJune 1, 1998. With permission from study participants, pancre-atic cancer was confirmed through physicians’ review of thenurses’ medical records. If permission was denied, we attemptedto confirm the self-reported cancer with an additional letter orphone call. We also searched the National Death Index toidentify deaths among the nonrespondents to each 2-year ques-tionnaire. The computerized National Death Index is a highlysensitive method for identifying death in this cohort (17). For alldeaths attributable to pancreatic cancer, we requested permissionfrom family members (subject to state regulation) to review themedical records. Pancreatic cancer was considered the cause ofdeath if the medical records or autopsy report confirmed fatalpancreatic cancer or if pancreatic cancer was listed as the un-derlying cause of death without another more plausible cause.

Statistical Analysis

We excluded participants who did not respond to the baselinequestionnaire or did not provide information on aspirin use onthe baseline questionnaire and all participants who reported abaseline history of cancer (with the exception of nonmelanomaskin cancer). We computed person-years of follow-up to the dateof diagnosis of pancreatic cancer, death from any cause, or theend of the study period (June 1, 1998), whichever occurred first.After these exclusions, 88 378 women were eligible for follow-up, and 1 475 262 person years were accrued. Among these88 378 women without cancer at baseline, we documented 161cases of pancreatic cancer during 18 years of follow-up. Becausethe use of analgesics might have been triggered by early symp-toms associated with pancreatic cancer, we excluded the first 2

Journal of the National Cancer Institute, Vol. 96, No. 1, January 7, 2004 ARTICLES 23



ownloaded from


years of follow-up, in secondary analyses, for all study partici-pants after their first report of analgesic use.

The primary analysis used incidence rates with person-yearsof follow-up in the denominator. We used relative risk as themeasure of association; relative risk was defined as the incidenceof pancreatic cancer among participants who reported regularuse of aspirin divided by the incidence among participantswithout such a report. We examined relative risks according toduration of regular aspirin use defined as duration before base-line (as reported on the 1980 questionnaire) plus years of regularuse after 1980. Regular use of aspirin was defined as the use oftwo or more tablets per week. In subanalyses, we examinedwhether regular use of five or more tablets of aspirin per weekinfluenced pancreatic cancer risk. Because data on the use ofsuch high doses of aspirin were not available before 1980(aspirin use was assessed as years of prior use of two or moretablets per week in 1980), the maximal report on duration ofregular aspirin intake, defined as five or more tablets per week,did not exceed 18 years in these analyses. To more accuratelyassess the influence of aspirin dose among relatively consistentaspirin users, we calculated the average aspirin dose amongwomen who reported any aspirin use on at least two of threeconsecutive biennial questionnaires (1980, 1982, and 1984) andcompared their incidence of pancreatic cancer to that amongwomen who reported no aspirin use on at least two of the samethree consecutive questionnaires.

Using Cox proportional hazards models, we adjusted forother potential risk factors for pancreatic cancer, including ahistory of diabetes and level of physical activity. We also con-ducted stratified analyses to determine whether the influence ofaspirin use was modified by body mass index and included aninteraction term of body mass index with dosage into the modelto test for statistical significance. The data conformed to pro-portional hazards assumptions, and all statistical tests were two-sided (� � .05).

RESULTS

During 18 years and 1 475 262 person-years of follow-up,161 women were diagnosed with pancreatic cancer. Baselinecharacteristics of the 88 378 women who completed the baselineaspirin questionnaire in 1980 and provided information on reg-ular aspirin use are shown in Table 1. At baseline, 43% of theparticipants reported any current, regular use of aspirin, definedas two or more aspirin tablets per week. Women who reported

current aspirin use were generally similar to women who did nottake aspirin. However, there was a higher prevalence of obesityand diabetes mellitus among participants who reported higherdoses or regular aspirin use.

Using data from the baseline 1980 report only, we examinedthe association between regular aspirin use, defined as two ormore tablets per day, and the risk of pancreatic cancer. Partici-pants who reported regular aspirin use at baseline, comparedwith those who reported using fewer than two tablets per week,experienced an increased risk of pancreatic cancer (RR � 1.43,95% CI � 1.05 to 1.95) after adjusting for age, cigarette smok-ing, history of diabetes mellitus, body mass index, and nonvig-orous physical activity.

We used reports that were updated biennially to examine theinfluence of current aspirin use on the risk of pancreatic cancer.Current regular aspirin use, defined as the use of two or moretablets per week at the beginning of each 2-year follow-up cycle,was not statistically significantly associated with the risk ofpancreatic cancer when compared with non-regular use (Table2). Adjustment for other known risk factors did not substantiallyalter this estimate (multivariable RR � 1.20, 95% CI � 0.87 to1.65). Increasing current doses of aspirin, compared with non-use, did not change these risks (Ptrend � .41). When we de-creased the number of categories of aspirin dosage to four byusing seven or more tablets per week as the highest category andsubjected the data to multivariable analysis, the risks associatedwith all dosage categories and the Ptrend remained statisticallynonsignificant.

We also examined the influence of increasing duration ofregular aspirin use (defined as two or more tablets per week) onthe risk of pancreatic cancer. Prolonged regular use of two ormore tablets per week appeared to increase the risk of pancreaticcancer (Ptrend � .01). Participants who reported the use of two ormore tablets per week for more than 20 years, compared withwomen who never regularly consumed aspirin at this dose,experienced a statistically significantly increased risk of pancre-atic cancer (multivariable RR � 1.58, 95% CI � 1.03 to 2.43).

We similarly assessed the influence of non-aspirin NSAIDuse on the risk of pancreatic cancer. Women who reported anycurrent other NSAID use, compared with non-users, had a sta-tistically nonsignificant increased risk of pancreatic cancer(RR � 1.20, 95% CI � 0.79 to 1.80). Because of limitedstatistical power, we could not examine the association be-

Table 1. Baseline characteristics according to baseline aspirin use*

Characteristic

Baseline No. of aspirin tablets per week

0 0.5–1.5 2–5 6–14 �14

No. of individuals 50 569 7178 14 343 10 544 5744Age, y (SD) 46.6 (7.3) 47.3 (7.1) 46.0 (7.0) 47.4 (7.1) 48.7 (6.9)Height, inches (SD) 64.4 (3.2) 64.4 (3.1) 64.4 (3.2) 64.5 (3.3) 64.5 (3.3)% with baseline body mass index �30† 10.0 10.3 10.3 12.7 16.8% with history of diabetes 2.5 1.9 1.9 2.6 3.1% current smokers 28.5 28.4 30.0 30.8 30.7Avg. No. (SD) of pack-years of smoking‡ 5.0 (8.0) 4.8 (7.8) 5.3 (8.1) 5.4 (8.4) 5.8 (8.8)Total physical activity, h/weekdays only (SD) 6.4 (3.9) 6.6 (3.8) 6.8 (3.9) 6.9 (3.9) 6.9 (3.9)

*Age was standardized according to eight categories (�44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, �75 y) as of 1980. Data are means (standard deviation[SD]), unless otherwise indicated.

†Body mass index is weight (in kilograms) divided by the square of height (in meters).‡Pack-years were calculated for former and current smokers only.




ownloaded from


tween either dose or duration of other NSAID use and pan-creatic cancer risk in this cohort.

To evaluate potential confounding by NSAID use, we re-stricted our study population, in separate analyses, to womenwho did not report any NSAID use in 1980 and on any subse-quent questionnaire to obtain a referent group that would includeonly non-users of any NSAIDs. Again, the relationship betweenlong-term regular aspirin use (two or more tablets per week) andthe risk of pancreatic cancer was not substantially altered (datanot shown).

Reported doses of aspirin use on any questionnaire includeda mixture of inconsistent and consistent users. Thus, to assess theinfluence of long-term, consistent use of aspirin more accurately,we focused our analyses on women who reported regular aspirinuse on at least two of three consecutive biennial questionnairesand compared their pancreatic cancer incidence from 1984through 1998 with that of women who were non-users on at leasttwo of the same three consecutive questionnaires (Table 3).Among aspirin users, we calculated an average dose of aspirin

based on the responses in 1980, 1982, and 1984 questionnaires.We observed a monotonic increase in the risk of pancreaticcancer with increasing aspirin dose, compared with consistentnon-use of aspirin (Ptrend � .02). Women with consistent use of14 or more tablets per week had the highest risk of pancreaticcancer (multivariable RR � 1.86, 95% CI � 1.03 to 3.35).

We further considered the possibility that changing our def-inition of regular aspirin use to use of five or more tablets perweek may influence our results and examined whether regularuse of five or more tablets of aspirin per week influenced the riskof pancreatic cancer. We used reports that were updated bien-nially (except 1986) to determine that current use of five or moreaspirin tablets per week was not statistically significantly asso-ciated with risk compared with use of fewer aspirin tablets(Table 4). However, increasing duration of regular use of thisdose of aspirin was associated with a statistically significantincrease in the risk of pancreatic cancer (Ptrend � .01). Partici-pants who regularly consumed five or more tablets of aspirin perweek for more than 10 years, compared with women who never

Table 2. Association between updated regular aspirin use (defined as use of �2 tablets per week) and incident pancreatic cancer*

Aspirin use No. of cases Total person-years Age-adjusted RR (95% CI) Multivariable RR (95% CI)

UseNon-regular user† 96 969 187 1.0 (referent) 1.0 (referent)Regular use (�2 tablets per wk) 65 506 075 1.26 (0.92 to 1.73) 1.20 (0.87 to 1.65)

Current aspirin use per wk0 tablets 57 642 553 1.0 (referent) 1.0 (referent)1–3 tablets 49 451 610 1.26 (0.86 to 1.85) 1.26 (0.85 to 1.85)4–6 tablets 18 135 816 1.50 (0.88 to 2.54) 1.41 (0.82 to 2.40)7–13 tablets 30 160 661 1.71 (1.10 to 2.65) 1.65 (1.05 to 2.59)�14 tablets 7 78 441 0.99 (0.45 to 2.20) 0.86 (0.39 to 1.89)

Ptrend‡ .17 .41Duration of regular use (�2 tablets per wk):

0 y 57 679 689 1.0 (referent) 1.0 (referent)1–5 y 24 260 997 0.95 (0.59 to 1.54) 0.94 (0.58 to 1.53)6–10 y 23 174 630 1.21 (0.75 to 1.96) 1.19 (0.73 to 1.95)11–20 y 23 143 104 1.49 (0.91 to 2.45) 1.51 (0.92 to 2.48)�20 y 34 214 650 1.58 (1.03 to 2.42) 1.58 (1.03 to 2.43)

Ptrend‡ .01 .01

*The number of cases of pancreatic cancer is shown. RR � relative risk; CI � confidence interval. Multivariable risks from proportional hazards models areadjusted for age in years, follow-up cycle, history of diabetes (yes/no), smoking status in six categories (never, quit �15 years ago, quit �15 years ago and smoked�25 pack-years in past 15 years, quit �15 years ago and smoked �25 pack-years in past 15 years, current smoker who smoked �25 pack-years in past 15 years,and current smoker who smoked �25 pack-years in past 15 years), quintiles of nonvigorous physical activity in metabolic equivalents per week (i.e., the caloric needper kilogram of body weight per hour of activity, divided by the caloric need per kilogram per hour at rest), and 1976 body mass index in five categories (�21,21–22.9, 23–24.9, 25–28.9, and �29 kg/m2).

†Non-regular use is defined as use of any aspirin dose of fewer than two tablets per week.‡All statistical tests were two-sided.

Table 3. Association between average aspirin dose and incident pancreatic cancer among consistent users and consistent non-users of aspirin (1984–1998)*

Aspirin dose per wk No. of cases Person-years Age-adjusted RR (95% CI) Multivariable RR (95% CI)

0 tablets 25 214 375 1.0 (referent) 1.0 (referent)1–3 tablets 69 585 545 1.07 (0.68 to 1.70) 1.11 (0.70 to 1.76)4–6 tablets 17 111 125 1.31 (0.71 to 2.43) 1.29 (0.70 to 2.40)7–13 tablets 17 94 064 1.46 (0.78 to 2.72) 1.41 (0.76 to 2.61)�14 tablets 20 75 574 2.02 (1.12 to 3.65) 1.86 (1.03 to 3.35)

Ptrend† .01 .02

*We calculated the average aspirin dose among women who reported any aspirin use on at least two of three consecutive biennial questionnaires (1980, 1982,1984); these analyses were based on 83 976 women and 149 incident pancreatic cancer cases that occurred between 1984 and 1998. RR � relative risk; CI �confidence interval. Multivariable risks from proportional hazards models are adjusted for age in years, follow-up cycle, history of diabetes (yes/no), smoking statusin six categories (never, quit �15 years ago, quit �15 years ago and smoked �25 pack-years in past 15 years, quit �15 years ago and smoked �25 pack-yearsin past 15 years, current smoker who smoked �25 pack-years in past 15 years, current smoker who smoked �25 pack-years in past 15 years), nonvigorous physicalactivity in metabolic equivalents per week (i.e., the caloric need per kilogram of body weight per hour of activity, divided by the caloric need per kilogram per hourat rest), in quintiles, and 1976 body mass index in five categories (�21, 21–22.9, 23–24.9, 25–28.9, and �29 kg/m2).

†All statistical tests were two-sided.




ownloaded from


regularly consumed five or more tablets of aspirin per week, hadan increased risk of pancreatic cancer (RR � 1.75, 95% CI �1.18 to 2.60). Because data on the use of such high doses ofaspirin were not available before 1980 (aspirin use was assessedas years of prior use of two or more tablets per week in 1980),we could not examine additional categories of use beyond 18years.

In previous analyses of this cohort, we observed that otherpancreatic cancer risk factors varied in effect by body massindex (16,18). We therefore examined the influence of bodymass index at baseline on the association between updatedregular aspirin use (two or more tablets per week) and the risk ofpancreatic cancer. Among women of normal weight (body massindex �25 kg/m2), we observed no association between regularaspirin use and the risk of pancreatic cancer (RR � 0.84, 95%CI � 0.54 to 1.30). However, among women whose body massindex was 25 kg/m2 or greater, the risk of pancreatic cancer wasgreater for current regular aspirin users (RR � 1.78, 95% CI �1.08 to 2.94) than for non-regular users.

Because the use of aspirin as an analgesic may have beentriggered by early symptoms associated with undiagnosed pan-creatic cancer, we analyzed the relationship between regularaspirin use as assessed in 1980 and the risk of pancreatic cancerfrom 1982 through 1998. Despite this 2-year latency period,current regular use of aspirin, compared with current non-regularuse of aspirin, was associated with a similar risk of pancreaticcancer (multivariable RR � 1.15, 95% CI � 0.83 to 1.59).Moreover, the risks associated with increasing aspirin doseamong consistent aspirin users (as defined in Table 4) alsoremained essentially unchanged when we excluded the first 2years of follow-up in these analyses: Women who consumed 14or more tablets per week, compared with consistent non-users,had an increased risk of pancreatic cancer (RR � 1.74, 95%CI � 0.89 to 3.42).

DISCUSSION

In this prospective cohort of women, participants who re-ported current use of two or more standard aspirin tablets per

week experienced a modest but not statistically significantlyincreased risk of pancreatic cancer compared with those whoreported less use. Furthermore, increasing duration of regularuse was associated with a statistically significant increase in risk.Among women who reported consistent aspirin use on at leasttwo of three consecutive biennial questionnaires, we observed astatistically significant increase in the risk of pancreatic cancerwith increasing aspirin dose compared with those who similarlyreported non-use.

Few studies have reported on the association between the useof aspirin and NSAIDs and the risk of pancreatic cancer. In acase–control study of 194 pancreatic cancer patients and 582age- and sex-matched control subjects, Menezes et al. (19)reported no statistically significant association between regularuse of aspirin and the risk of pancreatic cancer. The authors did,however, note a statistically nonsignificant association betweenincreased duration of use and pancreatic cancer risk (odds ratio[OR] � 1.21, 95% CI � 0.81 to 1.82 for 11 or more years ofuse). Coogan et al. (20) did not observe an increased risk ofpancreatic cancer associated with regular NSAID use (RR � 0.8,95% CI � 0.5 to 1.1) in their study of 504 case patients withpancreatic cancer and control subjects. In the largest case–control study, with 513 patients and 1535 control subjects,Langman et al. (21) found a statistically significantly increasedrisk of pancreatic cancer (OR � 1.49, 95% CI � 1.02 to 2.18)associated with aspirin and other NSAID use 13–36 monthsbefore diagnosis. In that study, which compared study subjectswith seven or more prescriptions with those with no prescrip-tions, data on aspirin and NSAID use were recorded in theprescription drug registry before the diagnosis of pancreaticcancer. Although the authors could not exclude the possibilitythat case patients were taking NSAIDs because of pain related tooccult, preclinical malignancy, the elevated risks persisted evenafter restricting analyses to prescriptions obtained more than 2years before diagnosis of pancreatic cancer.

Two prospective studies have reported an inverse associationbetween aspirin use and the risk of pancreatic cancer, but bothwere limited by a small number of case subjects and incompleteassessment of dosage and duration of aspirin and NSAID use.

Table 4. Association between updated regular aspirin use (defined as use of �5 tablets of aspirin per week) and incident pancreatic cancer*

Aspirin use No. of cases Total person-years Age-adjusted RR (95% CI) Multivariable RR (95% CI)

UseNon-regular use† 115 1 170 528 1.0 (referent) 1.0 (referent)Regular use (�5 tablets per wk) 46 304 735 1.36 (0.97 to 1.91) 1.28 (0.91 to 1.81)

Duration of regular use (�5 tablets per wk)‡0 y 81 945 311 1.0 (referent) 1.0 (referent)1–5 y 27 224 682 1.13 (0.73 to 1.75) 1.12 (0.72 to 1.74)6–10 y 16 118 690 1.12 (0.65 to 1.92) 1.10 (0.64 to 1.89)�10 y 37 185 264 1.80 (1.22 to 2.67) 1.75 (1.18 to 2.60)

Ptrend§ .01 .01

*The number of cases of pancreatic cancer is shown. RR � relative risk; CI � confidence interval. Multivariable risks from proportional hazards models areadjusted for age in years, follow-up cycle, history of diabetes (yes/no), smoking status in six categories (never, quit �15 years ago, quit �15 years ago and smoked�25 pack-years in past 15 years, quit �15 years ago and smoked �25 pack-years in past 15 years, current smoker who smoked �25 pack-years in past 15 years,current smoker who smoked �25 pack-years in past 15 years), nonvigorous physical activity in metabolic equivalents per week (i.e., the caloric need per kilogramof body weight per hour of activity, divided by the caloric need per kilogram per hour at rest), in quintiles, and 1976 body mass index in five categories (�21,21–22.9, 23–24.9, 25–28.9, and �29 kg/m2).

†Non-regular use is defined as use of any aspirin dose of fewer than five tablets per week.‡Duration of use was obtained from updated responses since 1980 and did not include years of prior use of two tablets or more per week in 1980. Thus, maximum

duration in this analysis was 18 years.§All statistical tests were two-sided.




ownloaded from


Using data from the National Health and Nutrition ExaminationSurvey, Schreinemachers and Everson (22) reported a decreasedrisk of pancreatic cancer (RR � 0.67, 95% CI � 0.33 to 1.67)among aspirin users from 30 patients with pancreatic canceraccrued over more than 12 years of follow-up. In a more recentprospective analysis of 80 patients with incident pancreaticcancer among women participants in the Iowa Women’s HealthStudy Cohort, Anderson et al. (10) reported a statistically sig-nificant inverse association between the risk of pancreatic cancerand any current aspirin use (RR � 0.57, 95% CI � 0.36 to 0.90),as well as a statistically significant trend with increasing doses ofaspirin (Ptrend � .005); no association was noted with the use ofother NSAIDs. However, in the Iowa cohort, data on updatedaspirin use and past use were not available. Consequently, theauthors could not assess the effect of longer durations of aspirinuse on the risk of pancreatic cancer.

Aspirin and NSAIDs may have several potential influenceson the pancreas that could affect pancreatic carcinogenesis.Some in vitro studies reported an inhibitory effect of thesecompounds on the growth of pancreatic cancer cell lines (8,23),although other investigations observed essentially no effect (24).In contrast, case reports and cohort studies have suggested thataspirin and NSAIDs are associated with an increased risk ofpancreatitis (25,26). Given the greater risk of pancreatic canceramong individuals with a history of pancreatitis (27), a subclin-ical chronic inflammation of the pancreas induced by long-termaspirin and NSAID use could elicit a modest increase in the riskof pancreatic cancer. Nonetheless, studies are needed to addressthe potential relationship between NSAID use and pancreaticcancer triggered through chronic pancreatitis, and these studieswould have to pay particular attention to the potential for indi-cation bias in such an association.

Another possible explanation for the observed increase inrisk after several years of aspirin use may be linked with itseffect on lipoxygenases. Until recently, NSAIDs, includingaspirin, were believed to influence carcinogenesis mainly byinhibiting cyclooxygenase 2. Cyclooxygenase 2 and lipoxy-genases metabolize polyunsaturated fatty acids and appear toaffect carcinogenesis. Recently, however, a better under-standing of the complex mechanisms involved in the chemo-preventive actions of NSAIDs is emerging, and other, not yetclearly defined, pathways have been hypothesized (28). Analternate pathway, as described by Ding et al. (29), forinstance, could be mediated through overexpression of5-lipoxygenase mRNA in pancreatic cancer. 5-Lipoxygenaseis a procarcinogenic lipoxygenase. Currently, the extent towhich NSAIDs inhibit anticarcinogenic lipoxygenases, asopposed to procarcinogenic lipoxygenases, is unclear. Anyimbalances induced by NSAIDs may impact differentially oncarcinogenesis in different tissues, especially in the long run,which could serve as an alternate explanation of the risk ofpancreatic cancer observed in our study.

In our cohort, the higher pancreatic cancer risk associatedwith aspirin use seemed to be confined to women with a higherbody mass index. Obesity and other symptoms of the metabolicsyndrome may be associated with low-grade inflammation, withone study (30) suggesting that inflammation precedes weightgain. Thus, obesity may have served as a marker for inflamma-tion in this cohort, a potential explanation for the observed effectmodification by body mass index.

The prospective nature of our studies precluded recall biasand the need for next-of-kin respondents. Moreover, to minimizemisclassification of exposure, we updated reports of drug intakebiennially. Self-reports of medication use are prone to error, butwe would expect such measurement error to be random bynature; thus it should only bias our results toward the null.Furthermore, multiple measures of aspirin use minimize thepotential impact of such error and allow us to assess the effectsof duration of aspirin use on pancreatic cancer risk. In addition,consistent with the results of randomized clinical trials, we havereported an inverse association between aspirin use and colorec-tal cancer and adenoma in this cohort (14), which suggests wehave a reasonably valid measure of aspirin use. Finally, becauseidentification of deaths is highly accurate in this cohort (17),differential follow-up is unlikely.

The positive association between aspirin use and the risk ofpancreatic cancer could reflect analgesic use for the treatment ofoccult or preclinical malignancies. However, the increasing riskof pancreatic cancer with increasing duration of use, particularlyafter more than 20 years, makes this explanation unlikely. More-over, our findings remained largely unchanged after we excludedthe first 2 years of follow-up.

In summary, our findings do not support a protective effect ofanalgesics use on the risk of pancreatic cancer. Rather, aspirinappears to increase the risk of pancreatic cancer after extendedperiods of use. This report supports current efforts examining thepotential differential effects of these agents in different tissues.Risks and benefits associated with the use of aspirin have to beweighed carefully in any recommendations made by health careproviders.

REFERENCES

(1) Jemal A, Murray T, Samuels A, Ghafoor A, Ward E, Thun MJ. Cancerstatistics, 2003. CA Cancer J Clin 2003;53:5–26.

(2) Thun MJ, Henley SJ, Patrono C. Nonsteroidal anti-inflammatory drugs asanticancer agents: mechanistic, pharmacologic, and clinical issues. J NatlCancer Inst 2002;94:252–66.

(3) Dannenberg AJ, Altorki NK, Boyle JO, Dang C, Howe LR, Weksler BB,et al. Cyclo-oxygenase 2: a pharmacological target for the prevention ofcancer. Lancet Oncol 2001;2:544–51.

(4) Vane JR, Bakhle YS, Botting RM. Cyclooxygenases 1 and 2. Annu RevPharmacol Toxicol 1998;38:97–120.

(5) Smith WL, Garavito RM, DeWitt DL. Prostaglandin endoperoxide Hsynthases (cyclooxygenases)-1 and -2. J Biol Chem 1996;271:33157–60.

(6) Tanabe T, Tohnai N. Cyclooxygenase isozymes and their gene structuresand expression. Prostaglandins Other Lipid Mediat 2002;68–69:95–114.

(7) Morrow JD, Roberts LJ. Lipid-derived autoacids. In: Hardman JG, LimbirdLE, editors. The pharmacologic basis of therapeutics. New York (NY):McGraw-Hill; 2001. p. 669–85.

(8) Perugini RA, McDade TP, Vittimberga FJ Jr, Duffy AJ, Callery MP.Sodium salicylate inhibits proliferation and induces G1 cell cycle arrest inhuman pancreatic cancer cell lines. J Gastrointest Surg 2000;4:24–32,discussion 32–3.

(9) Kokawa A, Kondo H, Gotoda T, Ono H, Saito D, Nakadaira S, et al.Increased expression of cyclooxygenase-2 in human pancreatic neoplasmsand potential for chemoprevention by cyclooxygenase inhibitors. Cancer2001;91:333–8.

(10) Anderson KE, Johnson TW, Lazovich D, Folsom AR. Association betweennonsteroidal anti-inflammatory drug use and the incidence of pancreaticcancer. J Natl Cancer Inst 2002;94:1168–71.

(11) Willett WC, Browne ML, Bain C, Lipnick RJ, Stampfer MJ, Rosner B, etal. Relative weight and risk of breast cancer among premenopausal women.Am J Epidemiol 1985;122:731–40.




ownloaded from


(12) Stampfer MJ, Willett WC, Colditz GA, Rosner B, Speizer FE, HennekensCH. A prospective study of postmenopausal estrogen therapy and coronaryheart disease. N Engl J Med 1985;313:1044–9.

(13) Colditz GA, Manson JE, Hankinson SE. The Nurses’ Health Study: 20-yearcontribution to the understanding of health among women. J WomensHealth 1997;6:49–62.

(14) Giovannucci E, Egan KM, Hunter DJ, Stampfer MJ, Colditz GA, WillettWC, et al. Aspirin and the risk of colorectal cancer in women. N EnglJ Med 1995;333:609–14.

(15) Fuchs CS, Colditz GA, Stampfer MJ, Giovannucci EL, Hunter DJ, RimmEB, et al. A prospective study of cigarette smoking and the risk ofpancreatic cancer. Arch Intern Med 1996;156:2255–60.

(16) Michaud DS, Giovannucci E, Willett WC, Colditz GA, Stampfer MJ,Fuchs CS. Physical activity, obesity, height, and the risk of pancreaticcancer. JAMA 2001;286:921–9.

(17) Stampfer MJ, Willett WC, Speizer FE, Dysert DC, Lipnick R, Rosner B, etal. Test of the National Death Index. Am J Epidemiol 1984;119:837–9.

(18) Michaud DS, Liu S, Giovannucci E, Willett WC, Colditz GA, Fuchs CS.Dietary sugar, glycemic load, and pancreatic cancer risk in a prospectivestudy. J Natl Cancer Inst 2002;94:1293–300.

(19) Menezes RJ, Huber KR, Mahoney MC, Moysich KB. Regular use ofaspirin and pancreatic cancer risk. BMC Public Health 2002;2:18.

(20) Coogan PF, Rosenberg L, Palmer JR, Strom BL, Zauber AG, Stolley PD,et al. Nonsteroidal anti-inflammatory drugs and risk of digestive cancers atsites other than the large bowel. Cancer Epidemiol Biomarkers Prev 2000;9:119–23.

(21) Langman MJ, Cheng KK, Gilman EA, Lancashire RJ. Effect of anti-inflammatory drugs on overall risk of common cancer: case-control studyin general practice research database. BMJ 2000;320:1642–6.

(22) Schreinemachers DM, Everson RB. Aspirin use and lung, colon, and breastcancer incidence in a prospective study. Epidemiology 1994;5:138–46.

(23) Molina MA, Sitja-Arnau M, Lemoine MG, Frazier ML, Sinicrope FA.Increased cyclooxygenase-2 expression in human pancreatic carcinomas

and cell lines: growth inhibition by nonsteroidal anti-inflammatory drugs.Cancer Res 1999;59:4356–62.

(24) Takahashi M, Furukawa F, Toyoda K, Sato H, Hasegawa R, Imaida K, et al.Effects of various prostaglandin synthesis inhibitors on pancreatic carcinogen-esis in hamsters after initiation with N-nitrosobis(2-oxopropyl)amine. Carci-nogenesis 1990;11:393–5.

(25) Blomgren KB, Sundstrom A, Steineck G, Genell S, Sjostedt S, Wiholm BE.A Swedish case-control network for studies of drug-induced morbidity–acute pancreatitis. Eur J Clin Pharmacol 2002;58:275–83.

(26) O’Brien WM, Bagby GF. Rare adverse reactions to nonsteroidal antiin-flammatory drugs. J Rheumatol 1985;12:13–20.

(27) Malka D, Hammel P, Maire F, Rufat P, Madeira I, Pessione F, et al. Riskof pancreatic adenocarcinoma in chronic pancreatitis. Gut 2002;51:849–52.

(28) Shureiqi I, Lippman SM. Lipoxygenase modulation to reverse carcinogen-esis. Cancer Res 2001;61:6307–12.

(29) Ding XZ, Kuszynski CA, El-Metwally TH, Adrian TE. Lipoxygenaseinhibition induced apoptosis, morphological changes, and carbonic anhy-drase expression in human pancreatic cancer cells. Biochem Biophys ResCommun 1999;266:392–9.

(30) Engstrom G, Hedblad B, Stavenow L, Lind P, Janzon L, Lindgarde F.Inflammation-sensitive plasma proteins are associated with future weightgain. Diabetes 2003;52:2097–101.

NOTES

Supported by Public Health Service grants CA/ES62984, CA87969,CA40356, CA55075, and CA86102 from the National Cancer Institute, NationalInstitutes of Health, Department of Health and Human Services.

We are indebted to the participants of the Nurses’ Health Study for theircontinuing outstanding dedication to the study. We thank Karen Corsano fortechnical assistance.

Manuscript received June 5, 2003; revised October 16, 2003; acceptedNovember 5, 2003.




ownloaded from


2013 cancer de pancreas y aas articulo original

Documents