Cancer Causes & Control

, Volume 17, Issue 7, pp 871–888

Aspirin and cancer risk: an updated quantitative review to 2005

Authors

    • Istituto di Ricerche Farmacologiche “Mario Negri”
  • Silvano Gallus
    • Istituto di Ricerche Farmacologiche “Mario Negri”
  • Carlo La Vecchia
    • Istituto di Ricerche Farmacologiche “Mario Negri”
    • Istituto di Statistica Medica e BiometriaUniversità degli Studi di Milano
Review Article

DOI: 10.1007/s10552-006-0033-7

Cite this article as:
Bosetti, C., Gallus, S. & La Vecchia, C. Cancer Causes Control (2006) 17: 871. doi:10.1007/s10552-006-0033-7

Abstract

Aspirin has been associated to a reduced risk of colorectal, and possibly of a few other common cancers. Epidemiological studies on aspirin and cancer risk published up to December 2005 have been reviewed, and pooled relative risks (RR) for several cancers have been provided. Besides a reduction in the risk of cancer of the colorectum (RR = 0.71, 95% confidence interval, CI: 0.67–0.75), there is evidence—although more limited, and mainly from case–control studies—that aspirin has a favourable effect on cancers of the esophagus (RR = 0.72, 95% CI: 0.62–0.84), stomach (RR = 0.84, 95% CI: 0.76–0.93), breast (RR = 0.91, 95% CI: 0.88–0.95), ovary (RR = 0.89, 95% CI: 0.78–1.02) and lung (RR = 0.94, 95% CI: 0.89–1.00). The role of aspirin on other cancers, such as pancreatic, prostate, bladder cancer, and non-Hodgkins’ lymphomas is less clear, and an increase of risk has been suggested for kidney cancer. For most cancer sites, however, significant heterogeneity between studies, and particularly across study design, was found, with a reduction in risk generally stronger in case–control than in cohort studies. Further, notwithstanding the large amount of epidemiological evidence, substantial uncertainties remain about the proper aspirin dose and duration of treatment.

Keywords

AspirinEpidemiologyNeoplasmReviewRisk factors

Introduction

Evidence of a protective role of aspirin on the risk of colorectal and a few other common cancers has been accumulated since the late 1980s [13]. A possible target of the cancer chemopreventive effect of aspirin and other NSAIDs is the inhibition of cyclooxygenase (COX), the enzymes responsible for the synthesis of prostaglandins. Indeed, COX (in particular the isoform COX-2) has been reported to be abnormally expressed in many cancer cell lines, and has been implicated in the process of carcinogenesis, tumor growth, apoptosis and angiogenesis [47].

Epidemiological studies on the association between aspirin and cancer risk published before 2001 have been previously reviewed, with a particular focus on colorectal, breast and ovarian cancers [8]. Since then, several articles have been published on the role of aspirin on several common cancers [9]. Thus, epidemiologic evidence from studies published up to 2001 is only briefly summarized in this paper, while papers published between January 2001 and December 2005 are described in detail. An overall quantitative estimate of the risk from all case–control and cohort studies published to date on aspirin use and cancer risk is also provided.

Materials and methods

The studies included in this quantitative review were all original cohort and case–control investigations on cancer including information on aspirin and other NSAID use published up to December 2005. They were identified through searches of the MEDLINE database, using the keywords “aspirin,” “NSAID,” “neoplasms,” “risk factors,” “case–control studies,” and “cohort studies.” Papers were also searched among those quoted as references in the retrieved studies. Articles reporting estimates for all NSAIDs combined only were not considered in the present review. A total of 82 papers were identified, including 46 case–control studies, and 33 cohort ones. The main characteristics and results of each study are described in Tables 17. These included the first author and year of publication, the total numbers of cases, the relative risk (RR) and corresponding 95% confidence intervals (CI) for regular aspirin use (or ever use, if the former was not given), and a brief comment.

To provide a quantitative overall estimate of the RR for various common cancers in relation to aspirin use, the RR and the 95% CI for regular (or ever) aspirin use were abstracted from the articles reviewed. Whenever available, estimates adjusted for multiple confounding factors were used; when RR and 95% CI were not given, they were calculated from tabular data. The pooled RR (and corresponding 95% CI) for each neoplasm, separately for case–control and cohort studies, were then calculated as a weighted average of the each study RR [10], with a weight proportional to the study precision (i.e., to the inverse of the variance of the log (RR); Table 8). The pooled estimates did not include the results from selected cohorts of subjects with rheumatoid arthritis [1113] and users of low-dose aspirin [14]. Tests for heterogeneity were used to evaluate the consistency of findings between studies and across study designs. Pooled estimates from a random-effects model, and corresponding tests of heterogeneity, gave similar results.

Results

Colon and rectal cancer

About 20 case–control and cohort studies have been published on aspirin and colorectal cancer [1536] (Table 1), the most frequent cancer among non-smokers in western countries [37].
Table 1

Main findings of epidemiological studies on aspirin and colorectal cancer

Study

No. cases

RR

95% CI

Comment

Case–control studies

Kune et al. [16]

715

0.53

0.40–0.71

Current vs. non-current use

Rosenberg et al. [17]

1,326

0.50

0.40–0.80

Regular use for ≥3 months

Suh et al. [18]

830

0.3–0.4

 

Use ≥2/day. Significant trend with dose

Muscat et al. [19]

511

0.64a

0.42–0.97

Protection unrelated to the indication. Significant trend with duration in men

0.32b

0.18–0.57

Peleg et al. [20]

216

0.25

0.09–0.73

Use in the four previous years >131 g. Significant trend with dose

Reeves et al. [21]

184

0.79

0.45–1.36

RR = 0.4, significant, for other NSAIDs

La Vecchia et al. [22]

1,357

0.70

0.50–1.00

Significant trend with dose

Rosenberg et al. [23]

1,201

0.70

0.50–0.90

Continued use. Significant protection for other NSAIDs

Friedman et al. [24]

1,993

0.70

0.60–0.80

RR = 0.60 for recent use. Similar protection, for other NSAIDs

Neugut et al. [25]

256

0.35

0.17–0.73

 

Juarranz et al. [26]

196

0.98

0.90–1.00

RR for a continuous increment. RR = 0.30, significant, for other NSAIDs

Sansbury et al. [27]

643

0.47

0.27–0.80

RR = 0.49, significant, for any NSAIDs

Cohort studies

Paganini-Hill et al. [28]

181

1.50

1.10–2.20

Daily use

Thun et al. [29, 30]

1,388

0.60

0.34–1.01

ACS/CPS II. ≥16 times/month for ≥1 year. Significant trend with frequency but not duration

Schreinemachers and Everson [31]

169

0.85

0.63–1.15

NHANES I. Incidence

Giovannucci et al. [32]

251

0.68

0.52–0.92

Male Health Professional. Student significant trend with the length of follow-up

Giovannucci et al. [33]

331

0.56

0.36–0.90

Nurses’ Health Study. Reduced risk for ≥20 years of use

García Rodríguez and Huerta Alvarez [34]

2,002

0.90

0.80–1.10

Nested case–control study. RR = 0.6, significant, for 300 mg/day for ≥6 months. RR = 0.7, significant, for other NSAIDs

Ratnasinghe et al. [35]

193

1.07

0.71–1.60

NHANES I and II. Mortality

Chan et al. [36]

962

0.77

0.67–0.88

Nurses’ Health Study. Regular use. Significant dose- and duration-response relations. RR = 0.47 for ≥14 aspirins/week and for ≥20 years of use

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; ACS: American Cancer Society; CPS: Cancer Prevention Study; and NHANES: National Health and Nutrition Examination Survey

aMen; bWomen; cUpdate of Giovannucci et al. [33]

A case–control study from Melbourne, Australia [16], on 715 colorectal cancer cases, first suggested a protective effect of aspirin, reporting a RR of 0.58 in males and of 0.49 in females who used aspirin. Some protection was also provided by other NSAIDs. A larger study was conducted in a network of American hospitals [17], based on 1,326 cases of colorectal cancer, with detailed data on drug use and taking into account several potential confounding factors. It found a RR of 0.5 for current aspirin users, but no association for ex-users. Eight additional case–control studies [1825] conducted in the USA, Great Britain and Italy, using either hospital or population controls, or within the framework of screening programs, reported RRs between 0.3 and 0.8.

With reference to data published after 2001, a case–control study from Spain [26], including 196 cases of colon cancer reported a reduced risk for aspirin (RR = 0.98 for a continuous increment), as well as for other NSAIDs (RR = 0.30, 95% CI: 0.1–0.98). Since this study did not provide risk estimates for users vs. non-users, it was not included in the pooled estimates. A case–control study from North Carolina [27], including 294 African-American and 349 White colon cancer cases, gave a RR of 0.47 (95% CI: 0.27–0.80) for aspirin use, and similar risk estimates for use of any NSAIDs (RR = 0.49) and of non-aspirin NSAIDs only (RR = 0.46).

Among cohort investigations, a study of elderly people (median age 73 years) found an increased risk of colorectal cancer [28]. The four main prospective studies, based on the American Cancer Society/Cancer Prevention Study II (ACS/CPS II) [29, 30], the National Health and Nutrition Examination Survey (NHANES I) [31], the Male Health Professionals [32], the Nurses’ Health Study [33, 36], and a nested case–control study from the UK General Practice Research Database [34] gave RRs between 0.6 and 0.9 for aspirin users compared to non-users, with inverse relations with frequency and duration of use. However, a recent analysis of mortality of the NHANES I and II cohorts up to the end of 1992 [35], including respectively 14,407 and 9,252 subjects, found a RR of 1.07 between aspirin use and colorectal cancer mortality.

Indirect epidemiological evidence on an inverse relation between aspirin and other NSAIDs on colorectal cancer risk has been also provided from three cohorts of patients with rheumatoid arthritis: a Swedish one [12], including 11,683 men and women, reported a RR of 0.63 for colon and of 0.72 for the rectal cancer; a cohort from Finland [11] on 46,101 patients found a RR of 0.90 for colon and of 0.60 for rectal cancer, and another cohort from Finland [13], including 9,469 patients, reported a RR of 0.62 for colorectal cancers combined.

With reference to time and dose of aspirin use, the Nurses’ Health Study [3336], based on a cohort of 82,911 women followed for 20 years, and including 962 cases of colorectal cancer, provided information on the relation between frequency and duration of aspirin use, and colorectal cancer risk. A reduced risk was observed with regular aspirin use (RR = 0.77, 95% CI: 0.67–0.88), the risk reduction being dose-related (RR = 1.10 for 0.5–1 aspirins/week, 0.89 for 2–5, 0.78 for 6–14, and 0.68 for ≥15). Moreover, the benefits became evident after more than five years of regular use, and were significant after 10 years. The data therefore suggest that there is a long latency between starting aspirin use and the subsequent protection against intestinal cancers. A nested case–control study from a British cohort, including 2,002 cases, showed that the risk of colorectal cancer was significantly reduced in subjects who had taken a daily dose of 300 mg aspirin for more than six months [34]. Likewise, no association between low-dose (up to 150 mg) aspirin use and incidence of colon and rectal cancer was reported in a cohort study from Denmark [14], linking the population-based Pharmacoepidemiologic Prescription Database with the Danish Cancer Registry, and including 29,470 individuals followed up to 9 years. The RR was 0.9 for colon cancer, based on 195 cases and 1.0 for rectal cancer, based on 104 cases.

Thus, the results of epidemiological studies conducted on different populations, using different methods and types of controls, and referring to more than 13,000 cases, indicate that (regular) aspirin use is associated with a reduction in the risk of colorectal cancer. The pooled RR estimate was 0.59 (95% CI: 0.54–0.64) from 11 case–control studies, 0.85 (95% CI: 0.78–0.92) from seven cohort ones, and 0.71 (95% CI: 0.67–0.75) from all studies combined (Table 8). The estimates from case–control and cohort studies were however heterogeneous, and significant heterogeneity was observed also within case–control and cohort studies.

Aspirin and NSAIDs have been shown to suppress the development of adenomatous polyps in patients with familial adenomatous polyposis [15, 38], and a few randomised trials [3942] showed that aspirin reduces the risk of colorectal adenomas in populations with history of colorectal cancer or adenomas. Similar protections against the risk of adenomas were found in various case–control and cohort studies [7]. Results from intervention studies on colorectal cancer are however inconclusive. Thus, the Physicians’ Health Study randomised trial [43, 44] found no effect of aspirin (325 mg) on invasive colorectal cancer incidence, although it was based on short treatment and limited follow-up. Similarly, the Women’s Health Study (WHS) randomised trial [45] did not report a protection for long-term aspirin use on colorectal cancer, at a dose of 100 mg/day.

The results of trials showing a beneficial effect of aspirin only in high-risk patients, and on a surrogate endpoint for colorectal cancer, suggest that it would be premature to recommend aspirin for primary prevention in the general population. Future studies are also needed to address the potential harmful effects of long-term aspirin use such as gastrointestinal bleeding [46].

Other digestive tract cancers

There are suggestions that the potential favourable role of aspirin and other NSAIDs on colorectal cancer might extend to other gastrointestinal cancers such as esophagus and stomach (Table 2) [30, 31, 35, 4758].
Table 2

Main findings of case–control studies of aspirin and digestive tract cancers other than colorectal

Study

No. cases

RR

95% CI

Comment

Esophagus

Case–control studies

Farrow et al. [47]

293a

0.37

0.24–0.58

≥1 tablet/week for ≥6 months. No relation with frequency and duration

221b

0.49

0.28–0.87

Sharp et al. [48]

159b

0.08

0.01–0.56

English center.

0.77

0.21–2.94

Scottish Center. Daily use for ≥1 month

Cohort studies

Thun et al. [30]

157

0.59

0.34–1.03

ACS/CPS II. ≥16 times/month for ≥1 year

Funkhouser and Sharp [49]

15

0.10

0.01–0.76

NHANES I. Incidence. Occasional use

Ratnasinghe et al. [35]

37

0.48

0.18–1.29

NHANES I and II. Mortality

Lindblad et al. [50]

909

0.93

0.76–1.15

Nested case–control study. Current use. RR = 0.76 for ≥3 years. Similar RR for non-aspirin NSAIDs

Stomach

Case–control studies

Farrow et al. [47]

261c

0.80

0.54–1.19

≥1 tablet/week for ≥6 months. No relation with frequency and duration. Stronger reduced risk for non-aspirin NSAIDs

368d

0.46

0.31–0.98

Zaridze et al. [51]

448

0.60

0.41–0.90

≥2 days/week for ≥6 months.

Akre et al. [52]

480

0.70

0.60–1.00

≥1 tablet/month. RR = 1.1 for non-aspirin NSAIDs

Cohort studies

Thun et al. [30]

266

0.53

0.34–0.81

ACS/CPS II. ≥16 times/month for ≥1 year

Schreinemachers and Everson [31]

39

0.93

0.49–1.74

NHANES I. Incidence

Ratnasinghe et al. [35]

48

0.82

0.38–1.81

NHANES I and II. Mortality

Lindblad et al. [50]

1,023

1.15

0.98–1.36

Nested case–control study. Current use. RR = 0.83 for non-aspirin NSAIDs

Pancreas

Case–control studies

Menezes et al. [53]

194

1.00

0.72–1.39

≥1 tablet/week for ≥6 months. No relation with dose nor duration

Cohort studies

Schreinemachers and Everson [31]

30

0.67

0.33–1.36

NHANES I. Incidence

Anderson et al. [54]

80

0.57

0.36–0.90

IWHS. RR = 0.40, significant, for ≥6 times/week. RR = 1.19 for non-aspirin NSAIDs

Schernhammer et al. [55]

161

1.20

0.87–1.65

Nurses’ HealthStudy. ≥2 tables/week. RR = 1.58 for >20 years of regular use

Jacobs et al. [56]

4,577

0.97

0.86–1.09

ACS/CPS-II. Daily use.

Ratnasinghe et al. [35]

78

0.87

0.42–1.77

NHANES I and II. Mortality

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; ACS: American Cancer Society; CPS: Cancer Prevention Study; NHANES: National Health and Nutrition Examination Survey; and IWHS: Iowa Women’s Health Study

aEsophageal adenocarcinomas; bSquamous-cell esophageal carcinomas; cGastric cardia adenocarcinomas; dNon-cardia gastric adenocarcinomas

A US case–control study of cancers of the esophagus [47], including 293 esophageal adenocarcinomas and 221 squamous-cell cancers, reported a protective effect for current aspirin use on adenocarcinomas (RR = 0.37) and squamous-cell cancers of the esophagus (RR = 0.49). Similar reduced risks were found for non-aspirin NSAIDs (RR = 0.57 and 0.33, respectively for the two histological types). No relation with frequency or with duration of use was however observed for both aspirin and non-aspirin NSAIDs. In a case–control study from England and Scotland [48], on 159 women with squamous-cell esophageal cancers, a significant inverse relation was seen in those taking aspirin at least once a month, but only in the English arm (RR = 0.08). Further, in an Italian case–control study [59] based on 956 patients with upper aerodigestive tract cancers—among which 225 were esophageal cancers—the RR was 0.89 for regular use and 0.33 (95% CI: 0.13–0.82) for use of ≥5 years.

The risk of esophageal cancer was reduced in the ACS/CPS II cohort (RR = 0.59 for aspirin use at least 16 times a month for at least one year) [30], in a follow-up of the NHANES I cohort for occasional aspirin users (RR = 0.1, based however on 15 cases only) [49], and in the mortality follow-up of the NHANES I and II cohorts (RR = 0.48) [35]. A nested case–control study based on the UK General Practice Research Database [50], including 909 subjects with esophageal cancer, reported a RR of 0.93 for current aspirin use, and of 0.76 for long-term (≥3 years) use. The corresponding RRs for use of non-aspirin NSAIDs were 0.95 and 0.82, respectively. The reduced risks were weakened in a two-year lag time analysis, and in subjects with no history of upper gastrointestinal disorders. A non-significant increase in esophageal cancer risk was reported in the Danish record linkage study of low-dose aspirin users (RR = 1.3, 95% CI: 0.9–1.9 based on 26 cases) [14]. No significant association was also observed in rheumatoid arthritis cohorts from Finland (RR = 0.94) [11] or Sweden (RR = 1.3) [12].

With reference to gastric cancer (Table 2), in a case–control study from the USA [47], including 261 gastric cardia adenocarcinomas and 368 other gastric adenocarcinomas, current aspirin users had a RR of 0.46 for non-cardia cancers and of 0.80 for cardia neoplasms, and there was no significant trend with frequency nor with duration of use. The reduction of risk was consistent for non-aspirin NSAIDs. In a case–control study from Russia [51], on 448 stomach cancer cases (92 gastric cardia and 356 other gastric cancers), the use of aspirin—and of other NSAIDs—was inversely related (RR = 0.60). Such an inverse relation seemed however limited to patients with non-cardia cancers (RR = 0.49) and with Helicobacter pylori (H. pylori) infection (RR = 0.31). In a case–control study from Sweden [52], comprising 567 cases, aspirin users had a RR of 0.7, with a significant inverse trend in risk with frequency of use (RR = 0.6 for ≥30 tablets/month). This reduction was similar for cardia and non-cardia cancers, and apparently more marked in patients positive for H. pylori.

As regards cohort studies, the risk of stomach cancer was reduced in the ACS/CPS II cohort, with a significant trend in risk for more frequent use (RR = 0.53 for aspirin use at least 16 times a month for at least one year) [30]. A RR of 0.93 for gastric cancer incidence was reported in the NHANES I cohort [31], and of 0.83 in a recent combined mortality analysis of the NHANES I and II cohorts [35]. A nested case–control study based on the UK General Practice Research Database [50], including 1,023 cases of gastric cancer, reported no reduction of risk both for current and for long-term (≥3 years) users (RR = 1.17 and 1.09, respectively). In the same study, non-aspirin NSAIDs gave a significant 17% reduced risk for current use and 15% for long-term use. The protection of non-aspirin NSAIDs was stronger in subjects with no history of upper gastrointestinal disorders. Similarly, no reduced incidence of stomach cancer was found in the Danish record linkage study of low-dose aspirin users (RR = 1.1, 95% CI: 0.8–1.3, based on 68 cases) [14]. The risk of stomach cancer was reduced, though not significantly, in Finnish [11] and Swedish [12] cohort studies of patients with rheumatoid arthritis (RR = 0.67 and 0.63, respectively).

Overall, the RR of esophageal cancer was 0.41 (95% CI: 0.29–0.57) from two case–control studies, 0.83 (95% CI: 0.70–0.98) from four cohort studies, and 0.72 (95% CI: 0.62–0.84) overall (Table 8). Corresponding figures for gastric cancer were 0.67 (95% CI: 0.56–0.80), 0.93 (95% CI: 0.82–1.05), and 0.84 (95% CI: 0.76–0.93), from three case–control, four cohort studies, and overall. For both esophageal and gastric cancer, there is evidence of a dose-dependent inverse relation with aspirin use, although the results from case–control and cohort studies were heterogeneous. The stronger inverse relation reported in case–control than in cohort studies may be partly attributed to potential greater bias of case–control studies. Since in fact aspirin and other NSAIDs may cause gastrointestinal bleeding, patients with early symptoms of undiagnosed esophageal and gastric cancer may be less prone to use these drugs. However, gastrointestinal bleeding caused by aspirin and other NSAIDs may have increased the frequency of medical examination, and thus the detection of early cancers otherwise undiagnosed, with the consequence that the inverse association between aspirin and these digestive tract cancers could have been obscured in both case–control and cohort studies. Further randomised trials on aspirin and cancer of the esophagus and stomach are thus warranted to confirm the results from observational studies, particularly in high-risk populations, such as subjects with H. pylori infection, gastro-esophageal reflux or Barrett’s esophagus.

With reference to pancreatic cancer (Table 2), a US case–control study [53] on 194 cases reported no association for aspirin (RR = 1.00), even for high dose and long duration of use. Among cohort studies, the NHANES I study [31] reported a non-significant reduced risk in ever aspirin users (RR = 0.67). An inverse risk was also reported in the Iowa Womens’ Health Study (IWHS) [54], where current users had a RR of 0.57, with a significant trend with frequency of use (RR = 0.40, 95% CI: 0.20–0.82, for users of six times per week). No relation was found for non-aspirin NSAID use. A slight reduction in pancreatic cancer mortality was reported in the follow-up of the NHANES I and II cohorts (RR = 0.87) [35]. Conversely, in the Nurses’ Health Study [55], based on more than 85,000 women, and including 161 incident pancreatic cancer cases, regular aspirin use was directly associated with pancreatic cancer risk: the RR was 1.20 for regular users (≥2 tablets per week), and 1.58 (95% CI: 1.03–2.43) for women who reported >20 years of regular aspirin use compared with women who were never regular users; a significant dose–risk relation was also reported for aspirin use on at least two of three consecutive biennial questionnaires. Even considering a 2-year lag time, use of aspirin was associated to an increased pancreatic cancer risk (RR = 1.15). However, the largest ACS/CPS II cohort [56], based on 4,577 deaths from pancreatic cancer, reported no association with aspirin use (RR = 0.97 for daily use and RR = 0.96, 95% CI: 0.69–1.55, for ≥20 years of daily use). No association with non-aspirin NSAIDs was found in a sub-cohort of subjects, including 311 pancreatic cancer cases. Similarly, the Danish record linkage study [14], including 62 pancreatic cancer cases, found no association in low-dose aspirin users (RR = 1.1, 95% CI: 0.8–1.3). In a Swedish cohort of patients with rheumatoid arthritis [12], a slightly reduced risk was reported (RR = 0.93, 95% CI: 0.6–1.2).

The studies on pancreatic cancer gave a pooled RR of 0.96 (95% CI: 0.92–1.01) overall, 1.00 (95% CI: 0.72–1.39) from one case–control study, and 0.96 (95% CI: 0.92–1.01) from five cohort studies (Table 8). The data are therefore too limited to draw any definitive conclusion on the association between aspirin and pancreatic cancer risk, although they allow to exclude a strong association.

Lung cancer

The relation between aspirin and lung cancer risk has been investigated in al least two case–control [60, 61] and six cohort studies [28, 30, 31, 35, 62, 63], whose results are given in Table 3.
Table 3

Main findings of case–control studies of aspirin and lung cancer

Study

No. cases

RR

95% CI

Comment

Case–control studies

Moysich et al. [60]

868

0.57

0.41–0.78

≥1 per week for ≥1 year

Muscat et al. [61]

1,038

0.84

0.62–1.14

Regular use. Risk reduction only in smokers.

Cohort studies

Paganini-Hill et al. [28]

111

1.35a

0.73–2.32

Daily use

0.29b

0.07–1.14

Thun et al.c [30]

nr

1.11a

0.98–1.25

ACS/CPS II. ≥16 times/month for ≥1 year

1.07b

0.88–1.30

Schreinemachers and Eversonc [31]

163

0.68

0.49–0.94

NHANES I study. Incidence.

Akhmedkhanov et al. [62]

81

0.66

0.34–1.28

Nested case–control study. ≥3 times per week for ≥6 months. RR = 0.68 for NSAIDs.

Holick et al. [63]

328

0.89

0.47–1.67

Health ProfessionalsStudy. Use ≥2/week

Ratnasinghe et al. [35]

410

0.81

0.62–1.07

NHANES I and II. Mortality

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; ACS: American Cancer Society; CPS: Cancer Prevention Study; and NHANES: National Health and Nutrition Examination Survey

aMen; bWomen; cRespiratory cancers

In a US population-based case–control study [60] including 868 cases, the RR for aspirin users were 0.57, with a stronger reduction for longer duration and higher frequency of use, both in small cell and non-small cell lung cancers. Another US case–control study [61], on 1,038 lung cancer cases, reported a RR of 0.84 for regular aspirin use, in the absence of any relation with number of tablets and frequency of use. The association was stronger in smokers and similar for various histological types, as well as for all NSAIDs combined.

As regards cohort studies, no association was reported in a cohort of elderly people (RR = 1.08 for both sexes combined) [28], and in the ACS/CPS II cohort (RR = 1.1) [30]. The NHANES I study [30] reported a significant reduction in lung cancer incidence (RR = 0.68). A small reduction in lung cancer mortality was reported in the follow-up of the NHANES I and II cohorts (RR = 0.81) [35]. A case–control study nested in the New YorkWomen’s Health Study cohort [62], including 81 female cases reported a non-significant RR of 0.66 for regular aspirin users. The RR was 0.39 (95% CI: 0.16–0.96) for non-small cells cancers. Similar results were found for ≥5 years of aspirin use (RR = 0.68). In the US HealthProfessionals cohort study [63] based on 328 male cases, the RRs were between 0.88 and 1.13 for different calendar years, and 0.89 overall, not significant. In a Danish study of low-dose aspirin [14], the RR for lung cancer incidence was 1.1 (95% CI: 0.9–1.2), on the basis of 274 cases.

The overall RR of lung cancer was 0.70 (95% CI: 0.56–0.88) from two case–control, 0.96 (95% CI: 0.91–1.02) from six cohort studies, and 0.94 (95% CI: 0.89–1.00) overall (Table 8). The pooled estimates from the two study designs were however heterogeneous, as well as the results of cohort studies. Thus, although there is some suggestion from case–control studies of a beneficial role of aspirin on lung cancer risk, the evidence remains inconsistent.

Breast and ovarian cancers

Studies on aspirin and the risk of breast cancer are shown in Table 4 [25, 28, 30, 31, 35, 6477]. Three case–control studies were published before 2001 [25, 64, 65], giving a pooled RR of 0.70 (95% CI: 0.61–0.81) [86]. Six cohort studies were published before 2001 [8], including the ACS/CPS II [30], the NHANES I study [31], the Nurses’ Health Study [71], and two other cohorts [70, 72], and their pooled RRs was 0.79 (95% CI: 0.59–1.06) [86]. All these cohort studies reported RRs for aspirin users below unity, with the exception of the Nurses’ Health Study (RR = 1.01) [71].
Table 4

Main findings of epidemiologic studies of aspirin and breast and ovarian cancer

Study

No. cases

RR

95% CI

Comment

Breast

Case–control studies

Harris et al. [64]

511

0.69

0.46–0.99

 

Neugut et al. [25]

252

0.80

0.35–1.80

 

Coogan et al. [65]

6,558

0.70

0.50–0.80

RR = 0.9 and 0.8 for other NSAIDs, in comparison with cancer and non-cancer patients

Cotterchio et al. [66]

3,133

0.73

0.61–0.87

RR = 0.76, significant, for all NSAIDs combined. RR = 0.68 for >8 years of NSAIDs use

Terry et al. [67]

1,442

0.80

0.66–0.97

Stronger protection for more frequent users (RR = 0.72) and in hormone receptor-positive tumors (RR = 0.74)

Swede et al. [68]

1,478

0.85

0.74–0.97

Regular use. RR = 0.74 for daily use. No trend with duration

Zhang et al.a [69]

7,006

0.86

0.64–1.16

Regular use. No trend with duration. RR = 0.74 for any NSAIDs and 0.85 for ibuprofen

Cohort studies

Friedman and Ury [70]

2

0.20

0.05–0.80

Computer-stored drug-dispensing data

Paganini-Hill et al. [28]

214

0.96

0.75–1.21

Daily use

Thun et al. [30]

nd

0.88

0.62–1.24

ACS/CPS II. ≥16 aspirins/month for ≥1 year

Schreinemachers and Everson [31]

147

0.70

0.50–0.96

NHANES I. Incidence

Egan et al. [71]

2,414

1.01

0.80–1.27

Nurses’ Health Study. No trend with frequency or duration of use

Harris et al. [72]

393

0.60

0.47–0.77

Similar protection for other NSAIDs

Johnson et al. [73]

938

0.82

0.71–0.95

IWHS, post-menopausal women. RR = 0.71 for more frequent use. RR = 0.98 for other NSAIDs

Harris et al. 2003 [74]

1,392

0.81

0.66–0.99

WHI. ≥5 years of use. Similar RR for all NSAIDs combined; stronger protection for ibuprofen

García-Rodríguez and González Pérez [75]

3,708

0.77

0.62–0.95

Nested case–control study. RR = 1.0 for non-aspirin NSAIDs and 0.76 for paracetamol. Greater reduced risk for low doses. No trend with duration

Ratnasinghe et al. [35]

131

0.82

0.49–1.36

NHANES I and II. Mortality

Jacobs et al. [76]

3,008

1.01

0.84–1.20

ACS/CPS-II Nutrition Cohort. ≥60 pills/month. RR = 1.06 for ibuprofen and 1.16 for other NSAIDs. No association with long-term regular use

Marshall et al. [77]

2,391

0.98

0.86–1.13

Daily use. No trend for frequency and duration among regular users RR = 0.89 in hormone receptor-positive cancers

Ovary

Case–control studies

Tzonou et al. [78]

189

0.51

0.26–1.02

Unspecified analgesics (mainly aspirin). Significant trend for frequency

Cramer et al. [79]

563

0.75

0.52–1.10

 

Tavani et al. [80]

749

0.93

0.53–1.62

No trend with frequency and duration

Rosenberg et al. [81]

780

0.8

0.5–1.2

RR = 0.5 for ≥5 years of use. RR = 0.5 for other NSAIDs

Moysich et al. [82]

547

1.00

0.73–1.39

RR = 0.87 for ≥7 aspirins/week

Akhmedkhanov et al. [83]

68

0.60

0.26–1.38

No trend with duration

Cohort studies

Fairfield et al. [84]

333

1.00

0.80–1.25

Nurses’ Health Study. No trend with dose nor duration. RR = 0.60, significant, for NSAIDS and 0.81 for paracetamol ≥5 days/month

Lacey et al. [85]

116

0.86

0.52–1.4

RR = 0.56 for more frequent use. No association for paracetamol and other NSAIDs

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; ACS: American Cancer Society; CPS: Cancer Prevention Study; NHANES: National Health and Nutrition Examination Survey; IWHS: Iowa Women’s Health Study; WHI: Women’s Health Initiative; and CPS: Cancer Prevention Study

aUpdate of Coogan et al. [65]

Among studies published more recently, a population-based case–control investigation from Ontario, Canada [66], including 3,133 cases of breast cancer reported a RR for daily use of aspirin for more than two months of 0.73. The RR was 0.79 (95% CI: 0.66–0.96) for non-aspirin NSAIDs and 0.76 (95% CI: 0.66–0.88) for all NSAIDs combined. The RR for NSAIDs was RR = 0.68 (95% CI: 0.54–0.86) for use longer than 8 years, while no meaningful trend was observed with time since first and last use, or age at first use. A population-based case–control study from the USA [67], including 1,442 cases, reported a RR of 0.80 for users of aspirin at least once per week, for six months or more. In the same study, the RRs for ever users were 0.91 (95% CI: 0.72–1.16) for ibuprofen, and 1.02 (95% CI: 0.80–1.31) for paracetamol. Stronger protections were observed for users of seven or more times per week (RR = 0.72), and for regular users, defined as users of at least four times per week for three months or more (RR = 0.73), while no relation was reported with duration of use. Further, the reduction in risk with aspirin was apparently restricted to women with hormone receptor-positive cancers (RR = 0.74). The latter results suggest that a possible mechanism involved in the protection of aspirin on breast cancer would be through suppression of the aromatase activity, i.e., reducing intramammary prostaglandins and thereby oestrogen production [87]. Further epidemiological data are however needed to confirm the interaction between aspirin use and hormone receptors. In a hospital-based case–control study from the Roswell Park Cancer Institute [68], including 1,478 breast cancer patients, the RR was 0.84 for regular aspirin use (≥1 tablet per week for ≥1 year), with a significant trend with increasing frequency (RR = 0.74, 95% CI: 0.59–0.92 for ≥7 tablets per week). The RR was 0.72 (95% CI: 0.53–0.97) for daily use of 10 or more years. An update of a US case–control study [64], including over 7,000 breast cancer cases [69], reported a weak inverse relation for regular aspirin use (RR = 0.86, 95% CI: 0.64–1.16), with no clear trend of reduction with increasing duration. Similar results were reported for any NSAIDs (RR = 0.74) and for ibuprofen (RR = 0.85). The association with aspirin was somewhat stronger for hormone receptor-positive cancers than for hormone receptor-negative cancers (RR = 0.74 and 0.81, respectively), although the estimates were not significantly heterogeneous.

With reference to cohort studies, the IWHS cohort [73], including 27,616 post-menopausal women, and 938 incident cases, reported a significant reduced breast cancer risk for ever-aspirin users (RR = 0.82) and a stronger association for use of six or more times per week (RR = 0.71, 95% CI: 0.58–0.87). No protection was reported for other NSAIDs use (RR = 0.98), even for more frequent use. The Women’s Health Initiative (WHI) Observational Study [74], based on 80,741 post-menopausal women, including 1,392 breast cancer cases, reported that regular NSAIDs use (two or more tablets per week) was associated to a 21% reduction of breast cancer risk for use lasting 5–9 years, and to a 28% reduction for 10 or more years. Consistent inverse associations were reported for regular use of aspirin (RR = 0.87 for 5–9 years, and 0.79 for 10 or more years), although they were less strong than for regular ibuprofen use (RR = 0.79 and 0.51, respectively). Paracetamol and low-dose (<100 mg) aspirin were unrelated to breast cancer risk. A case–control study nested in the UK General Practice Research Database cohort [75], including 3,708 breast cancer cases, reported that women taking aspirin for one year or more had a RR of 0.77. The results were similar for low-dose aspirin (75 mg, RR = 0.67). The RR was 1.00 for non-aspirin NSAIDs and 0.76 for paracetamol. No relation with duration of use was found both for aspirin and for other NSAIDs. A reduction in breast cancer mortality was reported in the follow-up of the NHANES I and II cohorts (RR = 0.82) [35]. The largest ACS/CPS II Nutrition Cohort [76], including 77,413 women and 3,008 breast cancer cases, however, reported no association between aspirin and breast cancer risk (RR = 1.01, for ≥60 pills per month). Corresponding RRs for ibuprofen and other NSAIDs were 1.06 and 1.16, respectively. A longer-duration of regular use (≥30 pills per month for ≥5 years) was unrelated to breast cancer risk both for aspirin (RR = 0.88) and for other NSAIDs (RR = 0.90), and was associated to an increased risk for ibuprofen (RR = 1.29). A Danish record linkage cohort study [14] reported a RR for breast cancer of 0.9 (95% CI: 0.8–1.1), based on 149 cases among low-dose aspirin users. Similarly, the California Teachers Study cohort [77], including 2,391 women with breast cases, reported that neither daily use of aspirin nor of any NSAID or paracetamol was associated with breast cancer risk (RR = 0.98, 1.09, and 0.99 respectively). Among regular aspirin users, no trend with duration or frequency of use was found. However, the RR was 0.80 (95% CI: 0.63–1.03) in hormone receptor-positive women using aspirin daily for at least 5 years.

Thus, epidemiological evidence on aspirin and breast cancer gave a pooled RR of 0.80 (95% CI: 0.73–0.87) from six case–control studies of 0.94 (95% CI: 0.90–0.98) from 12 cohort studies, and of 0.90 (95% CI: 0.87–0.94) overall (Table 8), in the presence of a significant heterogeneity between study design, and within cohort studies. Further quantification is needed, especially to clarify the long-term effects.

With reference to ovarian cancer, ovulation, and related inflammation, has been suggested to have a role in ovarian carcinogenesis [88, 89]. Epidemiological studies on aspirin use are still scanty, but there is an indication that aspirin may have a favourable effect on ovarian cancer (Table 4) [7885]. The pooled RR from six case–control studies published before 2001 [7883] was 0.84 (95% CI: 0.80–0.88) for aspirin users vs. non-users [8]. All studies reported RRs below unity, except an Italian [80] and an American [82] one.

The results of three cohort studies are however less convincing. The Nurses’ Health Study [84], based on 76,821 women followed up for 16 years, and including 333 cases of invasive epithelial ovarian cancer, reported a RR of 1.00 (95% CI: 0.80–1.25), and found no association for dose (RR = 0.98 for ≥15 tables per week) nor for duration (RR = 0.99 for ≥20 years) of use. No association was also found for paracetamol use (RR = 0.81 for ≥5 days per month). Conversely, a 40% reduction in risk was found for other NSAID use (RR = 0.60, 95% CI: 0.38–0.95), in the absence however of a dose–response relation. The Breast Cancer Detection Demonstration Project (BCDDP) follow-up [85], enrolling 61,431 women, and including 116 ovarian cancer cases, reported a non-significant reduction of ovarian cancer risk for regular aspirin use (RR = 0.86), but not for paracetamol (RR = 1.0) nor for other NSAIDs (RR = 1.0). The RR was 0.56, non-significant, for use of more than one aspirin per day for one year or more. A Danish record linkage cohort study [14], including 34 ovarian cancer cases among low-dose aspirin users, reported a RR of 1.1 (95% CI: 0.7–1.5).

The pooled RR of ovarian cancer for aspirin use was 0.82 (95% CI: 0.69–0.99) from six case–control studies, 0.98 (95% CI: 0.80–1.20) from two cohort ones, and 0.89 (95% CI: 0.78–1.02) overall (Table 8). The evidence is therefore too inconsistent to permit any definite conclusion.

Prostate cancer

The role of aspirin and other NSAIDs on prostate cancer risk is considered in Table 5 [25, 28, 30, 31, 9099]. The results of studies published before January 2003 were reviewed and summarized in a meta-analysis [100], which reported a pooled RR of 0.90 (95% CI: 0.82–0.99) for aspirin use, based on the combination of 10 studies, four retrospective [25, 9092], and six prospective [28, 30, 31, 9496] ones. However, the RR was 1.01 in retrospective and 0.85 in prospective studies, and the results for NSAIDs other than aspirin were inconsistent.
Table 5

Main findings of case–control studies of aspirin and prostate cancer

Study

No. cases

RR

95% CI

Comment

Case–control studies

Norrish et al. [90]

317

0.85

0.61–1.19

≥1 per week. RR = 0.88 for NSAIDs and 0.87 for non-aspirin NSAIDs. Stronger reductions for metastatic cancers.

Neugut et al. [25]

319

1.60

0.82–3.11

 

Menezes et al. [91]

1,096

1.08

0.87–1.35

 

Irani et al. [92]

639

0.95a

0.75–1.20

 

Bosetti et al. [93]

1,261

1.10

0.81–1.50

Use for ≥1 times week for ≥6 months. No trend with duration

Cohort studies

Paganini-Hill et al. [28]

149

0.95

0.60–1.51

Daily use

Thun et al.b [30]

nr

0.82

0.56–1.19

ACS/CPS II. ≥16 aspirins/month for ≥1 year

Schreinemachers and Everson [31]

123

0.95

0.66–1.35

NHANES I. Incidence

Leitzmann et al. [94]

2,479

1.04

0.86–1.26

Male Health Professional. ≥2 times/week reported in 4 consecutive questionnaires. RR = 0.73 for metastatic cancers

Habel et al. [95]

2,574

0.76

0.60–0.98

>6 tablets almost ever day

Perron et al. [96]

2,221

0.82

0.71–0.95

Nested case–control study. Daily dose of ≥80 mg for 8 years. RR = 0.72 for ≥4 years and ≥325 mg daily. No association with other NSAIDs

García Rodríguez and González Pérez [97]

2,183

0.70

0.61–0.79

Nested case–control study. Current use. No trend with duration. RR = 1.14, significant, for non-aspirin NSAIDs and 0.95 for paracetamol

Ratnasinghe et al. [35]

121

1.11

0.60–2.05

NHANES I and II. Mortality

Platz et al. [98]

141

0.76

0.54–1.16

Ever use. Lower reductions for current use. No trend with duration. Similar results for non-aspirin NSAIDs and paracetamol

Jacobs et al. [99]

4,853

0.95

0.82–1.10

ACS/CPS II Nutrition Cohort. ≥ 60 pills/month. Significant reduced risks with long-term regular use. RR = 0.92 for ibuprofen and 0.98 for other NSAIDs.

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; ACS: American Cancer Society; CPS: Cancer Prevention Study; and NHANES: National Health and Nutrition Examination Survey

aCrude estimate; bGenital cancers (including prostate and testis)

A least six studies have been subsequently published. A case–control study from Italy [93], on 1,261 cases of carcinoma of the prostate, reported a RR of 1.10 regular aspirin use, and found no relation either with duration or time since first use. In a nested case–control study from the UK General Practice Database cohort study [97], based on 2,183 prostate cases, the RR for current aspirin use was 0.70, in the absence of any trend with duration. A similar RR was reported for paracetamol, but not for other NSAIDs. In a US cohort study [98] based on 141 cases, the RR for ever-aspirin use was 0.76, again with no trend with duration of use. Similar RRs were reported for non-aspirin NSAIDs and paracetamol. The RR was 1.1 in the follow-up of the NHANES I and II cohorts (RR = 0.82) [34], and in a Danish cohort of low-dose aspirin users, based on 196 cases (95% CI: 1.0–1.3) [14]. Further, the ACS/CPS II Nutrition Cohort [98], including 70,144 men and 4,853 prostate cancer cases, showed no association for current aspirin use (RR = 0.95 for ≥60 pills/month, 95% CI: 0.82–1.11), as well as for ibuprofen (RR = 0.92) and other NSAIDs (RR = 0.98). However, long-duration regular aspirin use was associated with a reduced prostate cancer risk (RR = 0.85 for ≥5 years of use, 95% CI: 0.73–0.99). The association was stronger for advanced prostate cancer (RR = 0.64 for ≥5 years of use).

The pooled RR for prostate cancer was 1.02 (95% CI: 0.90–1.16) from five case–control studies, 0.97 (95% CI: 0.94–1.01) from 10 cohort studies, and 0.98 (95% CI: 0.95–1.01) overall (Table 8). Men taking aspirin on a regular basis may be more likely to have had frequent medical contacts and consequently prostate-specific antigen (PSA) measurements, and thus to have received a diagnosis of prostate cancer. Such a bias may well influence—to a variable extent—the results of cohort studies, too. The relation between aspirin use and prostate cancer, if any, remains inconclusive, but the data allow to exclude strong associations.

Bladder and kidney cancers

The main results of studies on aspirin and bladder cancer are reported in Table 6 [28, 30, 35, 101]. A population-based case–control study (1,514 cases) from Los Angeles County [101] reported a reduction in bladder cancer risk for exclusive aspirin users (RR = 0.85), with a trend for cumulative lifetime exposure (RR = 0.49 for ≥1,243 g). The RR was 0.81 for all NSAIDs combined, 1.03 for paracetamol, and 1.52 for phenacetin, an analgesic recognised as human urinary tract carcinogen [107]. In two older cohort studies [28, 30], the RR of bladder cancer for aspirin users was about 1.10. A strong increase in bladder cancer death was reported in the follow-up of the NHANES I and II cohorts (RR = 3.36) [34]. Similarly, in a Danish record linkage study [14], low-dose aspirin users had an increased bladder cancer incidence (RR = 1.2, 95% CI: 1.0–1.4, on the basis of 161 cases). The epidemiologic evidence on aspirin and bladder cancer is thus inconsistent, although a strong excess risk can now be excluded (pooled RR = 0.85, 95% CI: 0.66–1.09, from one case–control study, 1.23, 95% CI: 0.83–1.81, from three cohort studies, and 0.95, 95% CI: 0.77–1.17, overall; Table 8).
Table 6

Main findings of case–control studies of bladder and kidney cancer

Study

No. cases

RR

95% CI

Comment

Bladder

Case–control studies

Castelao et al. [101]

1,514

0.85

0.66–1.09

≥2 per week for ≥1 month. Trend with cumulative lifetime exposure. RR = 0.81, significant, for any analgesic and 1.03 for paracetamol

Cohort studies

Paganini-Hill et al. [28]

93

1.10

0.65–1.85

Daily use of aspirin for an undefined time

Schreinemachers and Everson [31]

35

1.06

0.54–2.09

NHANES I. Incidence

Ratnasinghe et al. [35]

40

3.36

0.60–2.05

NHANES I and II. Mortality

Kidney

Case–control studies

McLaughlin et al.c [102]

495

0.5a

0.2–1.0

More than 14 times per month for >36 months

1.8b

0.7–4.1

McCredie et al.a [103]

360

1.2

0.7–1.9

≥0.1 kg lifetime use

Mellemgaard et al.a [104]

368

1.4a

0.8–2.7

Ever use. RR = 3.1 and 4.0, respectively for men and women using >10,000 g. Similar risk for ever use of paracetamol

1.3b

0.7–2.6

McCredie et al. 1995c [105]

1,732

0.4

0.2–0.8

<0.1 kg lifetime use

1.1

0.9–1.3

≥0.1 kg lifetime use

Gago-Dominguez et al.c [106]

1,204

1.5

1.2–1.8

≥2 per week for ≥1 month. Trend with dose. Similar increased risk for other analgesics

Cohort studies

Paganini-Hill et al.c [28]

nr

6.3

2.0–20.0

Daily use

Schreinemachers and Everson [31]

32

0.60

0.29–1.24

NHANES I. Incidence

Ratnasinghe et al. [35]

37

2.27

0.93–5.54

NHANES I and II. Mortality

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; and NHANES: National Health and Nutrition Examination

aMen; bWomen; cRenal cell carcinoma only

At least eight studies investigated the relation between aspirin use and kidney cancer (Table 6) [28, 31, 35, 102106]. Among older case–control studies, four reported no overall association [102105]. A more recent study conducted in Los Angeles County [106], on a total of 1,204 renal cell carcinomas, found a 50% increased risk for regular (≥2 times a week for ≥1 month) aspirin use, with a significant trend in risk with dose (RR = 2.4 for ≥8 weekly doses). Similar results were found for other analgesics, including non-aspirin NSAIDs (RR = 1.4), paracetamol (RR = 1.7) and phenacetin (RR = 1.9). The increased risk for aspirin use was however limited to users of >325 mg per day.

Among cohort studies, one of elderly subjects [28] reported a significantly increased risk of kidney cancer for aspirin users (RR = 6.3); the NHANES I cohort found a non-significant reduced risk (RR = 0.60) [30]; and the follow-up of the NHANES I and II cohorts [35] found a reduction in kidney cancer mortality (RR = 0.82). Similarly, in a Danish record linkage study [14], low-dose aspirin users had an increased kidney cancer incidence (RR = 1.4, 95% CI: 1.1–1.7, on the basis of 67 cases).

Overall, the pooled RR for kidney cancer was 1.21 (95% CI: 1.07–1.36) from five case–control, 1.45 (95% CI: 0.87–2.40) from three cohort studies, and 1.22 (95% CI: 1.08–1.37) overall (Table 8). Thus, although based on a limited number of studies, the epidemiological evidence suggests an increased risk of kidney cancer for regular use of aspirin and other NSAIDs use. This increased risk may however be due to residual confounding by phenacetin use, which has been related to kidney cancer risk [102] and is likely to have been used in combination to aspirin and other NSAIDs.

Lymphatic and hematopoietic cancers

A few studies have investigated the relation between aspirin and the risk of non-Hodgkin's lymphomas (NHL, Table 7) [108110, 112, 113]. In a population-based case–control study from Los Angeles County [108], the RR of NHL was 1.63 for continuous use of aspirin or other NSAIDs for at least one month, and 1.96 for 13 months or longer. A case–control study conducted in upstate New York [109] on 376 cases of NHL reported a RR 2.38 (non-significant) for long-term use of aspirin; corresponding values were 1.39 for paracetamol, 5.64 for ibuprofen and 4.84 for other NSAIDs. Conversely, another case–control study from the USA [110], including 601 cases of NHL, reported RRs of 0.7 for aspirin and 1.0 for non-aspirin NSAIDs. The RR for long-term aspirin users was however 1.4. In a case–control study from the Roswell Park Cancer Institute [112] based on 625 cases, the RR of NHL for regular aspirin use was 0.82 in men and 0.93 in women. No evidence of trend was found for frequency, duration and cumulative exposure to aspirin. Regular use of paracetamol was associated to an increased NHL risk in women (RR = 1.71), but not in men (RR = 0.75).
Table 7

Main findings of case–control studies of aspirin and lymphatic and hematopoietic neoplasms

Study

No. cases

RR

95% CI

Comment

Case–control studies

Bernstein and Ross [108]

619

1.63

1.19–2.24

NHL. Continuous use of aspirin or other pain relieves for ≥1 month. RR = 1.96 for ≥13 months

Kato et al. [109]

376

2.38

0.75–7.56

NHL. >10 years of use. RR = 1.39 for paracetamol, 5.64 for ibuprofen and 4.84 for other NSAIDs

Zhang et al. [110]

601

0.7

0.4–1.2

NHL. No trend with duration RR = 1.4 for long duration. RR = 1.0 for non-aspirin NSAIDs.

Chang et al. [111]

565

0.60

0.42–0.85

HL. ≥2 tablets/week for ≥5 years. RR = 0.97 for non aspirin NSAIDs and 1.72 for paracetamol

Baker et al. [112]

625

0.82a

0.65–1.04

NHL. Regular use. RR = 0.57 and 1.71, for paracetamol use in men and women, respectively

0.93b

0.71–1.23

Cohort studies

Thun et al. [30]

nd

0.89a

0.66–1.20

Lymphatic and hematopoietic cancers. ACS/CPS II. ≥16 aspirins/month for ≥1 year

0.93b

0.65–1.33

Schreinemachers and Everson [31]

48

0.89

0.51–1.55

Lymphomas and leukemias. NHANES I. Incidence

49

0.67

0.34–1.31

Cerhan et al. [113]

131

1.21

0.81–1.87

NHL. IWHS. RR = 1.68 for exclusive use of aspirin. RR = 1.38 for non-aspirin NSAIDs

Kasum et al. [114]

81

0.45

0.27–0.75

Leukemias. IWHS. ≥2 times/week. RR = 1.31 for non-aspirin NSAIDs

Ratnasinghe et al. [35]

94

0.97

0.52–1.81

Lymphomas and leukemias. NHANES I and II. Mortality

63

1.08

0.58–2.01

RR: relative risk; CI: confidence interval; NSAIDs: non-steroidal anti-inflammatory drugs; NHL: non-Hodgkins’ lymphoma; HL: Hodgkins’ lymphoma; and IWHS: Iowa Women’s Health Study

aMen; bWomen

There was a reduced mortality from lymphatic and hematopoietic cancers in the ACS/CPS II cohort (RR = 0.89 for men and 0.93 for women for ≥16 aspirins per month) [30], and a lower incidence of lymphomas in the NHANES I cohort (RR = 0.89) [31]. In the IWHS cohort [113], ever users of aspirin had a 20% increased risk of NHL, and a higher risk was reported in exclusive aspirin users (RR = 1.68, 95% CI: 0.92–3.07). Corresponding values for other NSAIDs were 1.38 (95% CI: 0.97–1.95), and 2.21 (95% CI: 1.08–4.52), respectively. No association with mortality from lymphomas was reported in the NHANES I and II cohorts (RR = 0.97) [35], and with incidence of NHL in a Danish cohort of low-dose aspirin users [14]: the RR was 1.1 (95% CI: 0.9–1.5) on the basis of 57 cases.

The pooled RR for NHL was 0.98 (95% CI: 0.85–1.14) from four case–control studies, 1.08 (95% CI: 0.78–1.51) from two cohort studies, and 1.00 (95% CI: 0.88–1.14) from all studies combined (Table 8). The slight increased risk for NHL observed in some studies may be related to the immunomodulatory effects of aspirin and other NSAIDs, although a residual confounding by underlying chronic inflammation among patients with a long term anti-inflammatory drug users cannot be excluded [115].
Table 8

Pooled RR and 95%CI for aspirin users by cancer site

Cancer, study design

No. studies

Total number of cases

RRa

95% CI

Heterogeneity test p-value

Between studies

Between study designs

Colon and rectum

Case–control studiesb

11

9,232

0.59

0.54–0.64

0.008

<0.0001

Cohort studies

7

5,146

0.85

0.78–0.92

0.006

Esophagus

Case–control studies

2

643

0.41

0.29–0.57

0.26

0.0001

Cohort studies

4

1,118

0.83

0.70–0.98

0.18

Stomach

Case–control studies

3

1,557

0.67

0.56–0.80

0.42

0.0012

Cohort studies

4

1,376

0.93

0.82–1.05

0.002

Pancreas

Case–control studies

1

194

1.00

0.72–1.39

0.81

Cohort studies

5

4,926

0.96

0.92–1.01

0.04

Lung

Case–control studies

2

1,906

0.70

0.56–0.88

0.086

0.006

Cohort studiesc

6

1,003

0.96

0.91–1.02

0.012

Breast

Case–control studies

6

13,822

0.80

0.73–0.87

0.66

0.0012

Cohort studiesc

12

14,738

0.94

0.90–0.98

<0.0001

Ovary

Case–control studies

6

2,896

0.82

0.69–0.99

0.51

0.07

Cohort studies

2

449

0.98

0.80–1.20

0.59

Prostate

Case–control studies

5

3,632

1.02

0.90–1.16

0.44

0.48

Cohort studiesc

10

14,844

0.97

0.94–1.01

<0.0001

Bladder

Case–control studies

1

1,514

0.85

0.66–1.09

0.05

Cohort studies

3

168

1.23

0.83–1.81

0.21

Kidney

Case–control studies

5

3,796

1.21

1.07–1.36

0.004

0.003

Cohort studies

3

69

1.45

0.87–2.41

0.002

Non-Hodgkins’ lymphomas

Case–control studies

4

2,221

0.98

0.85–1.14

0.01

0.81

Cohort studiesc,d

2

142

1.08

0.78–1.51

0.39

aWeighted average of the each study estimate, with a weight proportional to the study precision, i.e., to the inverse of the variance of the estimate

bDoes not include Juarranz et al. [26]

cDoes not include Thun et al. [30]

dIncludes also Schreinemachers and Everson [31] and Ratnasinghe et al. [35]

Only four studies examined the relation between aspirin use and the risk of other lymphatic and hematopoietic neoplasms (Table 7) [111, 114]. A population-based case–control study from Connecticut [111] on 565 patients with Hodgkin's lymphomas reported a significantly lower risk for regular aspirin use (RR = 0.60). No association was found for non-aspirin NSAIDs (RR = 0.97), while a significant increased risk was reported for paracetamol (RR = 1.72). The IWHS cohort [114], including more than 28,000 postmenopausal women and 81 incident leukemias cases, reported a reduced RR for leukemias in regular users of aspirin (RR = 0.45, 95% CI: 0.25–0.79), but not of other NSAIDs (RR = 1.31). The incidence of leukemias was not significantly reduced in the NHANES I cohort (RR = 0.67) [31], and no association with leukemia mortality was found in the NHANES I and II cohorts (RR = 0.97) [35]. In a Danish cohort study [14], low-dose aspirin users had a RR of 1.3 (95% CI: 1.0–1.6), based on 69 cases of leukemias.

Discussion

Table 8 gives the pooled risk estimates from epidemiological studies on aspirin and several cancer sites. Besides a reduction in risk for cancer of the colorectum, there is evidence—although more limited, and mainly from case–control studies—that aspirin has a favourable effect on other digestive tract cancers, including those of the esophagus and stomach, on hormone-related neoplasms, such as the breast and ovary, and on lung cancer. The role of aspirin on other cancers, such as pancreatic, prostate and bladder cancers and NHL is less clear, and an increase of risk has been suggested for kidney cancer, possibly related to the associations with other or unspecified NSAIDs.

For most cancer sites, however, a significant heterogeneity between the estimates from various studies was found, and particularly between the results from case–control and cohort studies, with generally stronger reduction in risk in case–control studies than in cohort ones. Estimates from cohort studies are generally thought to be more reliable and valid, since these studies are generally less prone to information or selection bias. However, cohort studies may have less detailed information on drug use, and their often short follow-up do not allow to long-term effects of aspirin, whereas a detailed lifelong history of aspirin and other NSAIDs use can generally been obtained from case–control studies.

Further, notwithstanding the large amount of epidemiological evidence, substantial uncertainties remain about the proper aspirin dose and duration of treatment. For colorectal cancer it seems that doses over >300 mg/day [18, 20, 34] and many years of use [22, 33, 36] are needed to reduce the risk. Only a few studies, however, included information on dose and duration of use in relation to other cancers.

Moreover, most data on aspirin use and its possible chemopreventive effects are mainly based in observational studies, whereas the results from intervention studies, including the Physicians’ Health Study [33, 34] on colon cancer, and the WHS on several common neoplasms, are limited and inconclusive.

Acknowledgment

The authors thank Mrs I. Garimoldi for editorial assistance.

Copyright information

© Springer Science+Business Media B.V. 2006