Abstract
Prescription and over-the-counter medications are widely used in the United States and many western countries. More than two-thirds of women ages >45 years, who are at greatest risk for breast cancer, take prescription medication. In light of the ubiquitous nature of medication use and the fact that breast cancer remains the most common cancer in women, research on the role of medication use in breast cancer etiology is warranted. We summarize the epidemiologic evidence on the association between breast cancer risk and use of common medications, including antibiotics, antidepressants, statins, antihypertensives, and nonsteroidal anti-inflammatory drugs. Overall, there is little evidence that would implicate the use of antibiotics, antidepressants, statins, and antihypertensives in the etiology of breast cancer. Although several prospective studies and a randomized low-dose aspirin chemoprevention trial have not shown lower risk of breast cancer among aspirin users, most studies that have examined the potential chemoprotective effect of nonsteroidal anti-inflammatory drugs have shown significant risk reductions for regular and prolonged use of these drugs. The existing literature on the role of medication use in breast carcinogenesis is complicated. Interpretation of the evidence is hampered due to major methodologic differences across studies, including exposure assessment, exposure classification, and adjustment for potential confounding variables. These differences largely stem from the fact that the majority of articles on this topic represent secondary data analyses from studies with inadequate information on exposure or confounders. Thus, future epidemiologic studies specifically designed to study these ubiquitous and biologically plausible exposures are warranted. (Cancer Epidemiol Biomarkers Prev 2008;17(7):1564–95)
Introduction
Prescription and over-the-counter medications are very widely used in the United States and many western countries. A recent study of medication use in the ambulatory adult population of the United States revealed that 81% of participants have used at least one medication in the past week and that half of the sample reported to have taken at least one prescription medication. This survey also showed that women ages ≥65 years were the highest medication users; specifically, 12% of women in this age group took at least 10 different medications and 23% took at least 5 prescription drugs (1). More recent data from the Slone Survey (2) indicate that overall and prescription medication use has increased between 1999 and 2005. This study also reinforced earlier estimates that >90% of women ages ≥45 years reported any medication use. Further, prescription medication use for women ages 45 to 64 and ≥65 years was 68% and 82%, respectively. Thus, medication use in the United States represents a ubiquitous exposure. In light of the fact that breast cancer remains the most common cancer in women, a careful evaluation of the potential chemopreventive or carcinogenetic effects of common medications is warranted. In this review, we focus on commonly used medications that have been studied previously in epidemiologic studies of breast cancer. These groups of medications include antibiotics, antidepressants, statins, antihypertensives, and nonsteroidal anti-inflammatory drugs (NSAID).
Exposure Definition and Study Designs
The existing body of literature concerning the use of common medications and breast cancer risk is largely inconsistent. A primary reason for the divergent findings likely relates to the vast differences in methodologies employed in these studies. In addition to the obvious differences, such as study design (cohort studies versus case-control studies), these previous studies vary greatly with respect to exposure assessment, exposure classification, and adjustment for potential confounding variables. For instance, with respect to exposure assessment, many studies focused on NSAID use and breast cancer risk have only measured aspirin exposure but have no data on more recently introduced NSAIDs such as ibuprofen or selective cyclooxygenase-2 (COX-2) inhibitors. Thus, it is possible that women who do not report aspirin use but are in fact frequent ibuprofen users will be erroneously classified as “non-NSAID users,” because use of these newer drugs was not assessed in some studies. Further, using the existing research on antibiotic use and breast cancer risk as an example, there are great differences in exposure assessment. Some studies classify antibiotic use as crudely as “ever versus never,” whereas others have detailed information based on prescription data. Results from cohort studies might be difficult to interpret, as many studies rely on a single measurement of medication use, which does not take into account that medication use is subject to change over time. Further, many studies of medication use and breast cancer utilize large general practice databases, which improves exposure assessment but does not allow for adjustment for potential confounding variables, as these are generally not available in these data resources. Finally, it should be noted that the vast majority of existing studies represent so-called secondary data analyses, indicating that these various studies were not specifically designed to address the relationship between common medications and breast cancer risk. Rather, medication use was collected as a potential confounder or within the context of a medical history in which exposures or confounders is often absent. Although it is standard practice in epidemiologic research to analyze data for secondary associations, such studies are always methodologically inferior to those that were specifically designed to assess the link between specific medications and risk of breast cancer. Summarized below is the existing body of evidence of the associations between the use of common medications, such as antibiotics, antidepressants, statins, and NSAIDs, and breast cancer risk, preceded by a brief discussion of the biological mechanism by which these medications might influence risk.
Antibiotics and Breast Cancer Risk
Biological Mechanisms
A recent review of the biological mechanisms by which antibiotics may influence breast cancer risk suggests two main pathways: disruption of intestinal microflora and effect on immune and inflammatory function (3). Naturally occurring gut microflora have been shown to play a role in the conversion of phytochemicals derived from the consumption of plant-based food products into biologically active substances (4-6) suggested to be protective against cancer. For example, phytochemicals, such as lignans, can be converted by microflora to enterolactone (7), which has been correlated with reduced breast cancer risk (8, 9). Antibiotics could also theoretically decrease breast cancer risk by affecting the ability of microflora to modulate levels of circulating estrogens through deconjugation of bound estrogens in the gut, freeing them for reabsorption and circulation (10-13). However, the disruption of the microflora by antibiotics is not uniform and may vary by dose and specific drug formulation (8).
Breast cancer risk may also be mediated by the effect of antibiotics on the human immune system and inflammatory response. Numerous specific biological mechanisms have been suggested, but these remain largely speculative (3). Some antibiotics may have an anti-inflammatory effect by limiting the production of cytokines or a group of several proteins involved in the immune and inflammatory response (9). Inhibited cytokine production may be important in limiting estrogen synthesis in the peripheral fat (10, 11), potentially decreasing cancer risk. There is also limited evidence that some antibiotics may increase the production of prostaglandins or markers of the inflammatory response (3).
Summary of Existing Research
The potential role of antibiotic use in breast cancer etiology gained wide public attention after results from a recent large case-control study became available. In this study of 2,266 breast cancer patients and 7,953 controls who were enrolled in a nonprofit health plan, Lawlor et al. (14) were able to use computerized pharmacy records to assess exposure to antibiotic drugs. Results indicated that compared with women who never used antibiotics, women with the longest durations of antibiotic use had a 2-fold increase in breast cancer risk [odds ratio (OR), 2.07; 95% confidence interval (95% CI), 1.48-2.89]. Similar risk estimates were observed when nonusers were compared with women with the greatest number of antibiotic prescriptions (OR, 2.31; 95% CI, 1.69-3.15). Results were very similar for premenopausal and postmenopausal women and risk was increased for all subtypes of antibiotic drugs. These findings, which sparked considerable public concern about antibiotic use, are somewhat similar to those from a Finish cohort study (15) where ever use of antibiotics was associated with increased risk of breast cancer among premenopausal women [relative risk (RR), 1.74; 95% CI, 1.13-2.68] but not postmenopausal women (RR, 0.97; 95% CI, 0.59-1.58). Subsequent population-based (16) and nested case-control studies (17-19) did not report strong associations between antibiotic use and breast cancer risk. Most recently, Friedman et al. (12) conducted a 9-year follow-up study of >2 million women enrolled in the Kaiser Permanente Medical Care Program in northern California. They observed a modest risk elevation for women with the highest number of days using tetracyclines (RR, 1.23; 95% CI, 1.11-1.36) and an even more attenuated, nonsignificant estimate for macrolides (RR, 1.16; 95% CI, 0.98-1.36). Finally, in a case-case study, prolonged antibiotic use was not associated with tumor stage, grade, histology, or hormone receptor status (13).
As outlined in Table 1, there is little consensus on whether antibiotic use is associated with breast cancer risk. Any definitive conclusion is complicated by the fact that epidemiologic studies cannot distinguish between the potential carcinogenic effect of antibiotic drugs and the influence of the underlying conditions for which these drugs have been prescribed on breast cancer development.
Epidemiologic studies of the role of antibiotics use in breast cancer development
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Danielson et al. (15) Finland | 1973-1991 | Cohort study | 157 cases in Finnish Mobile Clinic Health Examination Survey identified via Finnish Cancer Registry | Self-report questionnaire | History of use of antibacterial treatment of bacteriuria vs none: 1.31 (0.95-1.81) | Age, region type, education, marital status, body mass index (BMI), parity, smoking, height, alcohol use, and screening positive for bacteriuria | |
9,304 cancer-free cohort members (total cohort 9,461) | Women ages <50 y, 1.74 (1.13-2.68); women ages >50 y, 0.97 (0.59-1.58) | ||||||
Lawlor et al. (14) USA | 1993-2001 | Population-based case-control study | Cases: 2,266 women enrolled in large health plan with primary invasive breast cancer identified from Surveillance, Epidemiology and End Results | Self-report questionnaire, health plan database | No appreciable difference by menopausal status; therefore all analyses combined | Age, length of health plan enrollment | Adjustment for following variables did not appreciably affect risk estimates: age at reference date, education, race, number of annual health-care visits, pharmacy co-pay status, age at menarche, parity, age at firth birth >30 y, BMI, family history of breast cancer, high breast density, hysterectomy, menopausal status, age at menopause ≥50 y, oral contraceptive use, and postmenopausal hormone replacement therapy (HRT) use |
Controls: 7,953 disease-free health plan members frequency matched 3:1 on age and sex | Highest number of days antibiotic use (≥1,001) vs none: 2.07 (1.48-2.89) | ||||||
Highest number of prescriptions filled (≥51) vs none: 2.31 (1.69-3.15) | |||||||
Wang et al. (16) Denmark | 1994-2003 | Registry-based study | Cases: 2,728 incident cases identified via population Hospital Discharge Registry | Epidemiologic prescription database | Highest number of prescriptions for antibiotics (>10) vs none: 1.00 (0.86-1.15) | Full sample: HRT use | |
Controls: 27,280 controls from population-based Civil Registration System matched 10:1 to cases | In women ages <70 y: 1.11 (0.93-1.32) | Women ages <70 y; HRT use, age at first birth, and parity | |||||
Weiss et al. (18) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases identified from General Practice Research Database | General practice database/electronic medical record | Highest number of days of antibiotic use (≥501) vs none: 1.2 (0.9-1.6) | Age, calendar year, BMI, alcohol intake, HRT use, NSAID use, prior benign breast disease, utilization of health services, time under observation | |
Controls: 20,000 frequency matched cancer-free controls | By indication vs none: respiratory infection 0.8 (0.7-1.0), urinary tract infection 0.9 (0.6-1.2), skin infection 1.2 (0.9-1.6), other infection 1.0 (0.8-1.3) | ||||||
Kaye and Jick (124) UK | 1987-2002 | Registry-based study | Cases: 1,268 cases identified from those with 6 y recorded medical history in General Practice Research Database | General practice database/electronic medical record | Highest number of days of use (≥501) vs none: 1.2 (0.6-2.4) | Risk estimates not appreciably changed when adjusted for following covariates: BMI, HRT use, history of benign proliferative breast disease, frequency of mammograms, frequency of visits to general practice | |
Controls: 6,291 cancer-free controls matched to cases up to 5:1 | |||||||
Didham et al. (125) New Zealand | 1998-2002 | Registry-based study | Cases: 700 cases (including 5 males) identified from General Practitioner Research Database | General practice database/electronic medical record | Ever prescription for any antibiotic vs none: 1.02 (1.0-1.05) | None | Analysis includes male breast cancers analyzed risk using conditional logistic regression |
Controls: 700 cancer-free controls matched 1:1 to cases on age, sex, semesters of available data | Ever penicillin vs none: 1.07 (1.02-1.13) | ||||||
Ever macrolide vs none: 0.90 (0.81-0.99) | |||||||
Ever tetracycline vs none: 1.06 (0.97-1.17) | |||||||
Velicer et al. (13) USA | 1993-2001 | Case only | 2,266 women with primary invasive breast cancer enrolled in Group Health Cooperative and identified through Surveillance, Epidemiology and End Results | Insurance plan prescription database and cost/utilization records, self-reported questionnaire | Antibiotic use ≥101 d vs none was not associated with tumor stage, grade, histology, or ER status | Age, length of enrollment | All OR >1: authors interpret as possible increase in less favorable tumor characteristics with antibiotic use |
Regional/distant vs local stage: 1.30 (0.93-1.81) | |||||||
Grade 4 vs 1: 1.39 (0.47-4.16) | |||||||
ER- vs ER+: 1.17 (0.79-1.75) | |||||||
Lobular vs ductal histology: 1.24 (0.79-1.96) | |||||||
Friedman et al. (12) USA | 1994-2003 | Registry-based study | Cohort: 2,130,829 female adult health subscribers | Insurance plan prescription database, subscriber surveys, medical record review | Any antibiotic use vs never 1.14 (1.10-1.18) | Hormone use | |
Cases: 18,521 women with incident invasive breast cancer | Use >1,000 d vs none: 1.17 (0.97-1.42) | ||||||
Use >100 d | |||||||
Tetracyclines: 1.23 (1.11-1.36), tetracyclines (excluding ever used macrolides): 1.14 (0.99-1.31) | |||||||
Macrolides: 1.16 (0.98-1.36), macrolides (excluding ever used tetracycline): 1.18 (0.93-1.49) | |||||||
Penicillin: 1.03 (0.94-1.13) |
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Danielson et al. (15) Finland | 1973-1991 | Cohort study | 157 cases in Finnish Mobile Clinic Health Examination Survey identified via Finnish Cancer Registry | Self-report questionnaire | History of use of antibacterial treatment of bacteriuria vs none: 1.31 (0.95-1.81) | Age, region type, education, marital status, body mass index (BMI), parity, smoking, height, alcohol use, and screening positive for bacteriuria | |
9,304 cancer-free cohort members (total cohort 9,461) | Women ages <50 y, 1.74 (1.13-2.68); women ages >50 y, 0.97 (0.59-1.58) | ||||||
Lawlor et al. (14) USA | 1993-2001 | Population-based case-control study | Cases: 2,266 women enrolled in large health plan with primary invasive breast cancer identified from Surveillance, Epidemiology and End Results | Self-report questionnaire, health plan database | No appreciable difference by menopausal status; therefore all analyses combined | Age, length of health plan enrollment | Adjustment for following variables did not appreciably affect risk estimates: age at reference date, education, race, number of annual health-care visits, pharmacy co-pay status, age at menarche, parity, age at firth birth >30 y, BMI, family history of breast cancer, high breast density, hysterectomy, menopausal status, age at menopause ≥50 y, oral contraceptive use, and postmenopausal hormone replacement therapy (HRT) use |
Controls: 7,953 disease-free health plan members frequency matched 3:1 on age and sex | Highest number of days antibiotic use (≥1,001) vs none: 2.07 (1.48-2.89) | ||||||
Highest number of prescriptions filled (≥51) vs none: 2.31 (1.69-3.15) | |||||||
Wang et al. (16) Denmark | 1994-2003 | Registry-based study | Cases: 2,728 incident cases identified via population Hospital Discharge Registry | Epidemiologic prescription database | Highest number of prescriptions for antibiotics (>10) vs none: 1.00 (0.86-1.15) | Full sample: HRT use | |
Controls: 27,280 controls from population-based Civil Registration System matched 10:1 to cases | In women ages <70 y: 1.11 (0.93-1.32) | Women ages <70 y; HRT use, age at first birth, and parity | |||||
Weiss et al. (18) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases identified from General Practice Research Database | General practice database/electronic medical record | Highest number of days of antibiotic use (≥501) vs none: 1.2 (0.9-1.6) | Age, calendar year, BMI, alcohol intake, HRT use, NSAID use, prior benign breast disease, utilization of health services, time under observation | |
Controls: 20,000 frequency matched cancer-free controls | By indication vs none: respiratory infection 0.8 (0.7-1.0), urinary tract infection 0.9 (0.6-1.2), skin infection 1.2 (0.9-1.6), other infection 1.0 (0.8-1.3) | ||||||
Kaye and Jick (124) UK | 1987-2002 | Registry-based study | Cases: 1,268 cases identified from those with 6 y recorded medical history in General Practice Research Database | General practice database/electronic medical record | Highest number of days of use (≥501) vs none: 1.2 (0.6-2.4) | Risk estimates not appreciably changed when adjusted for following covariates: BMI, HRT use, history of benign proliferative breast disease, frequency of mammograms, frequency of visits to general practice | |
Controls: 6,291 cancer-free controls matched to cases up to 5:1 | |||||||
Didham et al. (125) New Zealand | 1998-2002 | Registry-based study | Cases: 700 cases (including 5 males) identified from General Practitioner Research Database | General practice database/electronic medical record | Ever prescription for any antibiotic vs none: 1.02 (1.0-1.05) | None | Analysis includes male breast cancers analyzed risk using conditional logistic regression |
Controls: 700 cancer-free controls matched 1:1 to cases on age, sex, semesters of available data | Ever penicillin vs none: 1.07 (1.02-1.13) | ||||||
Ever macrolide vs none: 0.90 (0.81-0.99) | |||||||
Ever tetracycline vs none: 1.06 (0.97-1.17) | |||||||
Velicer et al. (13) USA | 1993-2001 | Case only | 2,266 women with primary invasive breast cancer enrolled in Group Health Cooperative and identified through Surveillance, Epidemiology and End Results | Insurance plan prescription database and cost/utilization records, self-reported questionnaire | Antibiotic use ≥101 d vs none was not associated with tumor stage, grade, histology, or ER status | Age, length of enrollment | All OR >1: authors interpret as possible increase in less favorable tumor characteristics with antibiotic use |
Regional/distant vs local stage: 1.30 (0.93-1.81) | |||||||
Grade 4 vs 1: 1.39 (0.47-4.16) | |||||||
ER- vs ER+: 1.17 (0.79-1.75) | |||||||
Lobular vs ductal histology: 1.24 (0.79-1.96) | |||||||
Friedman et al. (12) USA | 1994-2003 | Registry-based study | Cohort: 2,130,829 female adult health subscribers | Insurance plan prescription database, subscriber surveys, medical record review | Any antibiotic use vs never 1.14 (1.10-1.18) | Hormone use | |
Cases: 18,521 women with incident invasive breast cancer | Use >1,000 d vs none: 1.17 (0.97-1.42) | ||||||
Use >100 d | |||||||
Tetracyclines: 1.23 (1.11-1.36), tetracyclines (excluding ever used macrolides): 1.14 (0.99-1.31) | |||||||
Macrolides: 1.16 (0.98-1.36), macrolides (excluding ever used tetracycline): 1.18 (0.93-1.49) | |||||||
Penicillin: 1.03 (0.94-1.13) |
Antidepressant Use and Breast Cancer Risk
Biological Mechanisms
There are several tentative biological mechanisms by which antidepressants may play a role in breast cancer development. One frequently cited laboratory study found that the administration of antidepressants resulted in a significant increase in the development of mammary tumors in rodents (20). This positive association may be due to the structural similarities among common antidepressants and the cell growth regulating compound N,N-diethyl-2-[4-(phenylmethyl)phenoxy]ethanamine HCl. Tricyclic and selective serotonin reuptake inhibitors (SSRI) types of antidepressants have been shown to bind to the same intracellular histamine receptors associated with antiestrogen binding sites as N,N-diethyl-2-[4-(phenylmethyl)phenoxy]ethanamine HCl (20). However, the presumed effect of antidepressants on tumor growth was not replicated in subsequent in vitro studies of human breast tumor cell lines (21).
The cytochrome P450 enzyme system has been recognized as an important route of endogenous hormone metabolism, potentially affecting estrogen-dependent breast cancers. Myriad antidepressants have been shown to variably inhibit the cytochrome P450 system (22-25), increasing the availability of endogenous estrogens, thereby increasing the risk of breast cancer. Antidepressants are also thought to increase levels of prolactin (26, 27), itself a suspected breast tumor promoter. Finally, antidepressants may play a role in immune suppression by suppressing lymphocyte proliferation (28-30), suggesting an additional route for increased risk.
Summary of Existing Research
In a recent article, Lawlor et al. (14) conducted a systematic review of previous investigations aimed at exploring the association between antidepressant use and breast cancer risk. This review included seven relevant epidemiologic studies published until 2002: two prospective cohorts (31, 32), two retrospective cohort studies (15, 16), and three case-control studies (33-35). None of the case-control studies generated significant associations between antidepressant use and risk. One prospective cohort study (17) reported a significant increase in risk with use of any antidepressant at baseline only (RR, 1.75; 95% CI, 1.06-2.88). In contrast, a significant decrease in risk (OR, 0.50; 95% CI, 0.30-0.80) was found in one retrospective cohort study (15). In light of these inconsistent findings, the authors concluded in their review that the current epidemiologic evidence does not support an association between antidepressant use and breast cancer. A small case-control study, nested within a prescription database, which was not covered by the previous review, did not reveal an association between antidepressant use and risk (18).
Several epidemiologic studies have been published subsequent to the review article by Lawlor et al. (ref. 14; Table 2). Results from two population-based (19, 36) and one hospital-based (37) case-control studies did not show elevated breast cancer risk among antidepressant users. Similarly, two additional studies using general practice (38) and health-care plan (39) databases did not reveal significant associations with antidepressant use. In contrast, a large case-control study using the Saskatchewan Prescription Drug Plan (22) showed significant risk elevations for women who were prolonged users of certain genotoxic tricyclic antidepressants (amoxapine, clomipramime, and doxepin; OR, 2.39; 95% CI, 1.30-4.39) but not for nongenotoxic antidepressants (amitriptyline, maprotiline, and nortriptyline; OR, 1.02; 95% CI, 0.56-1.86). Genotoxicity assays were carried out using Drosophila melanogaster. Further, Fulton-Kehoe et al. (23) used a large health-care plan database and reported a modest increase in risk associated with ever use of amitriptyline (OR, 1.27; 95% CI, 1.10-1.47). However, no dose-response relationship was noted when number of prescriptions was considered, nor were risk elevations observed for tricyclic antidepressants or SSRI. Results from a small British cohort study did not reveal risk elevations for women who reported antidepressant use at ages 31 or 36 years (24). Finally, Chien et al. (25) reported results from a recent population-based case-control study where they observed significant risk increases for women with progesterone receptor (PgR)–negative tumors (OR, 1.8; 95% CI, 1.1-3.6) and estrogen receptor (ER)–positive/PgR-negative tumors (OR, 2.0; 95% CI, 1.1-3.8).
Epidemiologic studies of the role of antidepressants use in breast cancer development
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Weiss et al. (18) USA | 1988-1994 | Registry-based study | Cases: 95 cancer recurrences or 78 second primary cancer cases among adults with past history breast, colon, or melanoma | Insurance plan prescription database | Ever antidepressant use vs none and cancer recurrence: 0.97 (0.52-1.78) | Older study but not listed as reviewed by Lawler et al. | |
Controls: matched 5:1 on age, sex, original cancer type from a cohort of 1,467 patients | Ever antidepressant use vs none and second primary tumor: 0.94 (0.50-1.77) | Nest case-control design but no information given specifically about controls, only about entire cohort | |||||
Sharpe et al. (22) Canada | 1981-1995 | Registry-based study | Cases: 5,882 women with incident invasive breast cancer | Epidemiologic prescription database | Highest daily dose tricyclic antidepressant vs no use after 11-15 y: 2.02 (1.34-3.04) | Age, index date, tricyclic antidepressant use in other periods | Genotoxic tricyclic antidepressants: amoxapine, clomipramine, desipramine, doxepin, imipramine, trimipramine |
Controls: 23,517 women matched on age and sampling time | Genotoxic tricyclic antidepressants: 2.47 (1.37-4.40), nongenotoxic tricyclic antidepressants: 0.99 (0.49-1.99) | Nongenotoxic tricyclic antidepressants: amitryptyline, maprotiline, nortryptyline, protryptyline | |||||
Highest duration of use (71-100% of 11-15 y) | |||||||
Genotoxic tricyclic antidepressants: 2.39 (1.30-4.39), nongenotoxic tricyclic antidepressants: 1.02 (0.56-1.86) | |||||||
Moorman et al. (19) USA | 1996-2000 | Population-based case-control study | Cases: 938 cases of invasive breast cancer identified via a rapid case ascertainment system | In-person interviews | Any antidepressant use vs none: 1.0 (0.7-1.2) | Age and race | Adjustment for following variables did not appreciably effect risk estimates: age at menarche, menopausal status, family history of breast cancer in a first-degree relative, oral contraceptive use, HRT use, educational level, BMI, waist-to-hip ratio, alcohol consumption, and cigarette smoking |
Controls: 771 controls selected from DMV and Health Care Financing Administration | Any antidepressant ≥ 60 mo vs none: 1.0 (0.5-1.7) | ||||||
Tricyclic antidepressants ≥36 mo vs none: 0.7 (0.3-1.7) | |||||||
SSRI ≥36 mo vs none: 2.2 (0.8-6.3) | |||||||
Steingart et al. (36) Canada | 1996-1998 | Population-based case-control study | Cases: 3,133 female cases identified by the Ontario Cancer Registry | Self-administered questionnaire | Any regular antidepressant use vs none: 1.17 (1.01-1.36) | Age | Adjustment for following variables did not appreciably effect risk estimates: height, BMI, age at menarche, parity, age at menopause, oral contraceptive use, alcohol consumption, family history of breast cancer, and history of benign breast disease, clinical depression, or anxiety |
Controls: 3,062 population controls matched on age and sex | Highest duration of antidepressant use (≥9 y) vs none: 1.15 (0.78-1.69) | ||||||
Any SSRI use vs none: 1.33 (1.07-1.66) | |||||||
Sertraline: 1.58 (1.03-2.41) | |||||||
Paroxetine: 1.55 (1.00-2.40) | |||||||
Fluoxetine: 1.09 (0.81-1.49) | |||||||
Any tricyclic antidepressant use vs none: 1.12 (0.91-1.37) | |||||||
Amitriptyline: 1.10 (0.85-1.42) | |||||||
Imipramine: 0.88 (0.51-1.51) | |||||||
Doxepin: 1.21 (0.70-2.10) | |||||||
Any MAO-I use vs none: 0.87 (0.35-2.14) | |||||||
Gonzalez-Perez and Garcia Rodriguez (38) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | Past SSRI use vs none: 0.81 (0.67-1.00) | Age, calendar year, BMI, alcohol consumption, prior benign breast disease, depression, NSAID use, and HRT use | |
Database | SSRI duration >3 y vs none: 0.56 (0.27-1.18) | ||||||
Controls: 20,000 controls frequency matched on age, calendar year | SSRI high dose vs none: 0.95 (0.46-1.96) | ||||||
Past tricyclic antidepressant use vs none: 0.92 (0.80-1.05) | |||||||
Tricyclic antidepressant duration >3 y vs none: 0.83 (0.62-1.09) | |||||||
Tricyclic antidepressant high dose vs none: 1.11 (0.71-1.71) | |||||||
Haque et al. (39) USA | 1995-2000 | Registry-based study | 635 cases identified via Kaiser Permanente Southern California health plan cancer registry files | Large health plan database/electronic medical record | Ever used paroxetine vs never: 1.12 (0.96-1.31) | Age | Entire cohort use antidepressants, so controls were users of other medications |
Cohort: 109,004 women health plan members with a history of antidepressant use | Used paroxetine ≥2 y vs never: 0.90 (0.66-1.23) | ||||||
Coogan et al. (37) USA | 1988-2002 | Hospital-based case-control study | Cases: 2,138 cases of primary, invasive breast cancer identified via discharge summaries and pathology reports | In-person interviews | Regular SSRI use vs none: 1.1 (0.8-1.7) | Age, study center, year of interview, alcohol consumption, religion, family history of breast cancer, and race | |
Controls: 2,858 patients without cancer diagnoses frequency matched to cases on age, study center, and interview year | Paroxetine vs none: 0.8 (0.3-2.3) | ||||||
Regular SSRI duration ≥4 y vs none: 0.7 (0.4-1.5) | |||||||
Fluoxetine duration ≥4 y vs none: 0.9 (0.4-2.2) | |||||||
Sertraline duration ≥4 y vs none: 1.0 (0.3-4.1) | |||||||
Fulton-Kehoe et al. (23) USA | 1990-2001 | Population-based case-control study | Cases: 2,904 cases of primary invasive or in situ breast cancer in women enrolled in a large HMO, identified via Surveillance, Epidemiology and End Results | Health plan prescription database, self-administered questionnaire | Ever use antidepressants vs never: 1.04 (0.94-1.16) | Age, length of enrollment, calendar year, family history of breast cancer, parity/age at first birth, duration of HRT use, BMI, history of screening mammogram in 2 y before reference date | |
Controls: 14,396 disease-free controls matched 5:1 to cases on age, calendar year, and length of HMO membership | Any antidepressant, 51+ Rx: 1.12 (0.89-1.41) | ||||||
Ever tricyclic antidepressant vs none: 1.06 (0.94-1.19) | |||||||
Ever amitriptyline: 1.21 (1.03-1.41) | |||||||
Ever doxepine: 0.95 (0.79-1.13) | |||||||
Ever imipramine: 1.04 (0.84-1.29) | |||||||
Ever SSRI vs none: 0.98 (0.80-1.18) | |||||||
Any SSRI + Rx: 1.04 (0.73-1.48) | |||||||
Ever fluoxetine: 1.00 (0.80-1.25) | |||||||
Ever paroxetine: 1.00 (0.70-1.41) | |||||||
Ever sertraline: 1.16 (0.81-1.66) | |||||||
Lokugamage et al. (24) UK | 1946-2005 | Cohort | Cohort of 2,253 women followed from birth | In-person interviews | Use of an antidepressant at age 31 or 36 vs never use: 0.75 (0.27-2.05) | None | |
Cases: 83 women with incident breast cancer | |||||||
Chien et al. (25) USA | 1997-1999 | Population-based case-control study | Cases: 975 women age 65-79 y with primary invasive cancer | In-person interview | Ever antidepressant use vs never: 1.2 (0.9-1.6) | Age, year, county of residence | Adjustment for following variables did not appreciably effect risk estimates: race, income, marital status, education, time since last medical checkup, age at menarche, parity, age at first birth, type of menopause, age at menopause, duration of contraceptive use, menopausal hormone use, family history of breast cancer, tobacco smoking, alcohol consumption, BMI, and various medical conditions |
Controls: 1,007 women matched on age, year, and county of residence | Antidepressant use among +FHx: 0.4 (0.2-0.9), antidepressant use among -FHx: 1.5 (1.1-2.0) | ||||||
Antidepressant use among ER+: 1.2 (0.9-1.6), antidepressant use among ER-: 1.6 (1.0-2.8) | |||||||
Antidepressant use among PgR+: 1.1 (0.8-1.5), antidepressant use among PgR-: 1.7 (1.1-2.5) | |||||||
Ever tricyclic antidepressant use vs none: 1.2 (0.8-1.8), ever tricyclic antidepressant among +FHx: 0.5 (0.2-1.3), ever tricyclic antidepressant among -FHx: 1.5 (0.9-2.3) | |||||||
Ever SSRI use vs none: 1.2 (0.8-1.8), ever SSRI among +FHx: 0.4 (0.2-1.0), ever SSRI among -FHx: 1.4 (0.9-2.2) |
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Weiss et al. (18) USA | 1988-1994 | Registry-based study | Cases: 95 cancer recurrences or 78 second primary cancer cases among adults with past history breast, colon, or melanoma | Insurance plan prescription database | Ever antidepressant use vs none and cancer recurrence: 0.97 (0.52-1.78) | Older study but not listed as reviewed by Lawler et al. | |
Controls: matched 5:1 on age, sex, original cancer type from a cohort of 1,467 patients | Ever antidepressant use vs none and second primary tumor: 0.94 (0.50-1.77) | Nest case-control design but no information given specifically about controls, only about entire cohort | |||||
Sharpe et al. (22) Canada | 1981-1995 | Registry-based study | Cases: 5,882 women with incident invasive breast cancer | Epidemiologic prescription database | Highest daily dose tricyclic antidepressant vs no use after 11-15 y: 2.02 (1.34-3.04) | Age, index date, tricyclic antidepressant use in other periods | Genotoxic tricyclic antidepressants: amoxapine, clomipramine, desipramine, doxepin, imipramine, trimipramine |
Controls: 23,517 women matched on age and sampling time | Genotoxic tricyclic antidepressants: 2.47 (1.37-4.40), nongenotoxic tricyclic antidepressants: 0.99 (0.49-1.99) | Nongenotoxic tricyclic antidepressants: amitryptyline, maprotiline, nortryptyline, protryptyline | |||||
Highest duration of use (71-100% of 11-15 y) | |||||||
Genotoxic tricyclic antidepressants: 2.39 (1.30-4.39), nongenotoxic tricyclic antidepressants: 1.02 (0.56-1.86) | |||||||
Moorman et al. (19) USA | 1996-2000 | Population-based case-control study | Cases: 938 cases of invasive breast cancer identified via a rapid case ascertainment system | In-person interviews | Any antidepressant use vs none: 1.0 (0.7-1.2) | Age and race | Adjustment for following variables did not appreciably effect risk estimates: age at menarche, menopausal status, family history of breast cancer in a first-degree relative, oral contraceptive use, HRT use, educational level, BMI, waist-to-hip ratio, alcohol consumption, and cigarette smoking |
Controls: 771 controls selected from DMV and Health Care Financing Administration | Any antidepressant ≥ 60 mo vs none: 1.0 (0.5-1.7) | ||||||
Tricyclic antidepressants ≥36 mo vs none: 0.7 (0.3-1.7) | |||||||
SSRI ≥36 mo vs none: 2.2 (0.8-6.3) | |||||||
Steingart et al. (36) Canada | 1996-1998 | Population-based case-control study | Cases: 3,133 female cases identified by the Ontario Cancer Registry | Self-administered questionnaire | Any regular antidepressant use vs none: 1.17 (1.01-1.36) | Age | Adjustment for following variables did not appreciably effect risk estimates: height, BMI, age at menarche, parity, age at menopause, oral contraceptive use, alcohol consumption, family history of breast cancer, and history of benign breast disease, clinical depression, or anxiety |
Controls: 3,062 population controls matched on age and sex | Highest duration of antidepressant use (≥9 y) vs none: 1.15 (0.78-1.69) | ||||||
Any SSRI use vs none: 1.33 (1.07-1.66) | |||||||
Sertraline: 1.58 (1.03-2.41) | |||||||
Paroxetine: 1.55 (1.00-2.40) | |||||||
Fluoxetine: 1.09 (0.81-1.49) | |||||||
Any tricyclic antidepressant use vs none: 1.12 (0.91-1.37) | |||||||
Amitriptyline: 1.10 (0.85-1.42) | |||||||
Imipramine: 0.88 (0.51-1.51) | |||||||
Doxepin: 1.21 (0.70-2.10) | |||||||
Any MAO-I use vs none: 0.87 (0.35-2.14) | |||||||
Gonzalez-Perez and Garcia Rodriguez (38) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | Past SSRI use vs none: 0.81 (0.67-1.00) | Age, calendar year, BMI, alcohol consumption, prior benign breast disease, depression, NSAID use, and HRT use | |
Database | SSRI duration >3 y vs none: 0.56 (0.27-1.18) | ||||||
Controls: 20,000 controls frequency matched on age, calendar year | SSRI high dose vs none: 0.95 (0.46-1.96) | ||||||
Past tricyclic antidepressant use vs none: 0.92 (0.80-1.05) | |||||||
Tricyclic antidepressant duration >3 y vs none: 0.83 (0.62-1.09) | |||||||
Tricyclic antidepressant high dose vs none: 1.11 (0.71-1.71) | |||||||
Haque et al. (39) USA | 1995-2000 | Registry-based study | 635 cases identified via Kaiser Permanente Southern California health plan cancer registry files | Large health plan database/electronic medical record | Ever used paroxetine vs never: 1.12 (0.96-1.31) | Age | Entire cohort use antidepressants, so controls were users of other medications |
Cohort: 109,004 women health plan members with a history of antidepressant use | Used paroxetine ≥2 y vs never: 0.90 (0.66-1.23) | ||||||
Coogan et al. (37) USA | 1988-2002 | Hospital-based case-control study | Cases: 2,138 cases of primary, invasive breast cancer identified via discharge summaries and pathology reports | In-person interviews | Regular SSRI use vs none: 1.1 (0.8-1.7) | Age, study center, year of interview, alcohol consumption, religion, family history of breast cancer, and race | |
Controls: 2,858 patients without cancer diagnoses frequency matched to cases on age, study center, and interview year | Paroxetine vs none: 0.8 (0.3-2.3) | ||||||
Regular SSRI duration ≥4 y vs none: 0.7 (0.4-1.5) | |||||||
Fluoxetine duration ≥4 y vs none: 0.9 (0.4-2.2) | |||||||
Sertraline duration ≥4 y vs none: 1.0 (0.3-4.1) | |||||||
Fulton-Kehoe et al. (23) USA | 1990-2001 | Population-based case-control study | Cases: 2,904 cases of primary invasive or in situ breast cancer in women enrolled in a large HMO, identified via Surveillance, Epidemiology and End Results | Health plan prescription database, self-administered questionnaire | Ever use antidepressants vs never: 1.04 (0.94-1.16) | Age, length of enrollment, calendar year, family history of breast cancer, parity/age at first birth, duration of HRT use, BMI, history of screening mammogram in 2 y before reference date | |
Controls: 14,396 disease-free controls matched 5:1 to cases on age, calendar year, and length of HMO membership | Any antidepressant, 51+ Rx: 1.12 (0.89-1.41) | ||||||
Ever tricyclic antidepressant vs none: 1.06 (0.94-1.19) | |||||||
Ever amitriptyline: 1.21 (1.03-1.41) | |||||||
Ever doxepine: 0.95 (0.79-1.13) | |||||||
Ever imipramine: 1.04 (0.84-1.29) | |||||||
Ever SSRI vs none: 0.98 (0.80-1.18) | |||||||
Any SSRI + Rx: 1.04 (0.73-1.48) | |||||||
Ever fluoxetine: 1.00 (0.80-1.25) | |||||||
Ever paroxetine: 1.00 (0.70-1.41) | |||||||
Ever sertraline: 1.16 (0.81-1.66) | |||||||
Lokugamage et al. (24) UK | 1946-2005 | Cohort | Cohort of 2,253 women followed from birth | In-person interviews | Use of an antidepressant at age 31 or 36 vs never use: 0.75 (0.27-2.05) | None | |
Cases: 83 women with incident breast cancer | |||||||
Chien et al. (25) USA | 1997-1999 | Population-based case-control study | Cases: 975 women age 65-79 y with primary invasive cancer | In-person interview | Ever antidepressant use vs never: 1.2 (0.9-1.6) | Age, year, county of residence | Adjustment for following variables did not appreciably effect risk estimates: race, income, marital status, education, time since last medical checkup, age at menarche, parity, age at first birth, type of menopause, age at menopause, duration of contraceptive use, menopausal hormone use, family history of breast cancer, tobacco smoking, alcohol consumption, BMI, and various medical conditions |
Controls: 1,007 women matched on age, year, and county of residence | Antidepressant use among +FHx: 0.4 (0.2-0.9), antidepressant use among -FHx: 1.5 (1.1-2.0) | ||||||
Antidepressant use among ER+: 1.2 (0.9-1.6), antidepressant use among ER-: 1.6 (1.0-2.8) | |||||||
Antidepressant use among PgR+: 1.1 (0.8-1.5), antidepressant use among PgR-: 1.7 (1.1-2.5) | |||||||
Ever tricyclic antidepressant use vs none: 1.2 (0.8-1.8), ever tricyclic antidepressant among +FHx: 0.5 (0.2-1.3), ever tricyclic antidepressant among -FHx: 1.5 (0.9-2.3) | |||||||
Ever SSRI use vs none: 1.2 (0.8-1.8), ever SSRI among +FHx: 0.4 (0.2-1.0), ever SSRI among -FHx: 1.4 (0.9-2.2) |
Overall, these additional reports also do not provide strong evidence that would implicate antidepressant use in the etiology of breast cancer. More detailed analyses by hormone receptor status in existing data sets might be warranted.
Statin Drug Use and Breast Cancer Risk
Biological Mechanisms
There is considerable interest and controversy around whether statins may play a role in carcinogenesis. An early laboratory study suggested that lipid-lowering drugs cause cancer in rodents at amounts that would be comparable with clinically effective doses in humans (40). However, several studies published subsequently have called those findings into question. The best-studied route of action for statins appears to be their inhibition of 3-hydroxy-3-methylglutaryl coenzyme-A reductase, a key enzyme in the mevalonate pathway of cholesterol synthesis. Inhibition of 3-hydroxy-3-methylglutaryl coenzyme-A reductase thereby inhibits prenylation, a protein synthesis process that leads to cell signaling processes involved in cell proliferation (28, 41). Preclinical studies have shown that a variety of statins working through disruption of the mevalonate pathway decrease cell proliferation by promotion of G1 cell cycle arrest and apoptosis in breast cancer cell lines (29-31, 42). Statins have also been shown to decrease mammary tumor formation and metastasis in a mouse model (32).
Interest in the mevalonate synthesis as target for cancer therapies has grown with the observation that statins may show a synergistic effect with chemoradiation (43), chemotherapies (33, 34, 44), and COX-2 inhibitors (35). Independent of the mevalonate pathway, statins have been suggested to have anticancer properties through an anti-inflammatory effect and via inhibition of the proteasome (41).
Summary of the Existing Evidence
The association between statin use and breast cancer risk has been the subject in recent attention in the field of pharmacoepidemiology (Table 3). Many of these studies used prescription or health-care plan record databases. Results from these investigations have consistently not revealed strong associations between statin use and risk (45-53). Although findings from these geographically diverse investigations are consistent, they may have to be cautiously interpreted due to significant methodologic shortcomings such as lack of adjustment for confounders and crude exposure assessment (ever versus never) in many of these studies. Coogan et al. (54) reported findings from a hospital-based case-control study in which prolonged statin use was associated with 2-fold increase in breast cancer risk (OR, 2.1; 95% CI, 1.1-4.0). However, more detailed analyses revealed that this estimate was largely driven by women with in situ disease (OR, 3.4; 95% CI, 1.5-8.0) rather than by women with invasive breast cancer (OR, 1.5; 95% CI, 0.7-3.1). In a more recent report by these investigators, prolonged statin use was not significantly associated with breast cancer risk (55). These latter findings are consistent with those of a population-based case-control study where ever and prolonged statin use was not associated with excess risk (56). Further, analyses from two large cohort studies, the Nurses' Health Study (57) and the Women's Health Initiative Observational Study (58), did not reveal significant associations. In contrast, Cauley et al. (59) described results from a smaller cohort study where ever use of statin drugs was associated with a significant risk reduction (OR, 0.28; 95% CI, 0.09-0.86). However, this estimate was based on a very small number of exposed breast cancer patients (n = 6) and results should be interpreted cautiously. Finally, two recent meta-analyses on this topic did not provide evidence that statin use is linked to breast cancer risk (60, 61). Thus, considering this diverse and largely consistent body of evidence, it is unlikely that statin drug use is an important factor in breast cancer development.
Epidemiologic studies of the role of statin drug use in breast cancer development
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Peeters et al. (52) Denmark | 1991-1994 | Registry-based study | 6 cases were identified using a population-based prescription database and the Danish Cancer Registry | Epidemiologic prescription database | Use of statins vs none, standardized incidence ratio: 1.4 (0.5-3.1) | ||
1,882 patients in cohort, with 4,580 person-years of follow-up | |||||||
Jick et al. (46) Quebec | 1988-1994 | Registry-based study | Cases: 56 breast cancer cases were identified using computerized health databases of the Regie de l'Assurance-Maladie du Quebec | Computerized health record database | Ever used statin vs use of bile acid binding resins: 0.67 (0.33-1.38) | Age at index date, previous neoplasm, year of cohort entry, use of fibric acid, use of other lipid-reducing agents, and a comorbidity score | |
Controls: 560 cancer-free controls matched to cases (6,721 patients in cohort) | |||||||
Michels et al. (51) UK | 1992-1998 | Registry-based study | Cases: 224 incident invasive and in situ carcinomas from the General Practice Research Database | General practice database/electronic medical record | Current statin use vs none: 1.0 (0.6-1.6) | ||
Controls: 1,009 cancer-free matched controls | Past statin use vs none: 1.3 (0.6-2.8) | ||||||
Statin use >5 y vs none: 1.1 (0.4-3.0) | |||||||
Li et al. (80) USA | 1987-2001 | Hospital-based case-control study | Cases: 1,132 primary invasive and in situ breast cancer confirmed by pathology report | In-person interview | For all breast cancers, use of statins ≥ 3 y vs none: 2.1 (1.1-4.0) | Age, year of interview, study center, education, number of doctor visits two y before hospitalization, use of conjugated estrogens, HRT use, oral contraceptive use, religion, race, alcohol consumption, and BMI | |
Controls: 589 women with noncancer, non-statin-related conditions | For carcinoma in situ, use of statins ≥3 y vs none: 3.4 (1.5-8.0) | ||||||
For invasive breast cancer, use of statins ≥3 y vs none: 1.5 (0.7-3.1) | |||||||
Alshafie et al. (87) USA | 1992-2001 | Cohort study | 244 incident breast cancer cases confirmed by medical record and pathology report | Questionnaire and interviews | Ever used statins vs none: 0.28 (0.09-0.86) | Age and body weight | Adjustment for following variables did not appreciably effect risk estimates: HRT use, family history of breast cancer, mammography use, height, education, health status, age at menarche, age at first birth, parity, physical activity, and alcohol consumption |
7,284 cancer-free cohort members | Ever used nonstatin lipid-lowering drug vs none: 0.37 (0.14-0.99) | ||||||
Ever used any lipid-lowering drug vs none: 0.32 (0.15-0.68) | |||||||
Gonzalez-Perez et al. (45) Canada | 1989-1997 | Registry-based study | 879 incidence breast cancers identified through regional cancer registry | Computerized health record database | Ever used statins vs none: 1.09 (0.93-1.28) | None | |
Cancer-free cohort members (total cohort = 67,472) | Age ≤55 y and ever used statins vs none: 0.81 (0.53-1.24) | ||||||
Age >55 y and ever used statins vs none: 1.15 (0.97-1.37) | |||||||
Age >55 and ≥37 y HRT use vs none: 2.04 (1.20-3.46) | |||||||
Boudreau et al. (56) USA | 1997-1999 | Population-based case-control study | Cases: 975 primary, invasive cancers identified via tumor registry/Surveillance, Epidemiology and End Results | In-person interview | Ever used statins vs none: 0.9 (0.7-1.2) | Age, reference year, county of residence, use of antihypertensive medication | |
Controls: 1,007 cancer-free general population controls identified via Medicare/Medicaid lists | Current statin use >5 y vs none: 0.7 (0.4-1.0) | ||||||
Kaye and Jick (126) UK | 1990-2002 | Registry-based study | Cases: 3,224 incident cancer cases, including 698 breast cancers from the General Practice Research Database | General practice database/electronic medical record, self-administered questionnaire | Current statin use vs none: 0.9 (0.6-1.3) | Not specified for breast model, but considered BMI, smoking status, average general practice visit frequency during follow-up | |
Controls: 14,844 cancer-free matched controls | |||||||
Graaf et al. (127) The Netherlands | 1991-1998 | Registry-based study | Cases: 3,129 incident cancer cases, including 467 breast cancers from the PHARMO drug dispensing database system | Drug dispensing database linked to hospital discharge record | Ever statin use vs none: 1.07 (0.65-1.74) | Diabetes mellitus, prior hospitalizations, chronic disease score, chronic use of diuretics, ACEi, CCB, hormones, NSAID, and other lipid-lowering therapy | |
Controls: 16,976 cancer-free matched controls | |||||||
Olsen et al. (48) Denmark | 1989-2002 | Registry-based study | Cases: 22,512 incident cancer cases, including 3,141 breast cancer cases identified via Central Population Register, Epidemiologic Prescription Database, and Danish Cancer Registry | Epidemiologic prescription database | Ever statin use vs none: 1.02 (0.76-1.36) | Age, calendar period, NSAID, HRT, cardiovascular drugs | |
Controls: 334,754 men and women in general population, with 12,251 statin users | |||||||
Brueggemeier et al. (85) USA | 1994-2000 | Cohort study | Cases: 3,177 incident cases of breast cancer identified from self-report and medical record review | Self-administered questionnaire | Current statin use vs none: 0.91 (0.76-1.08) | Age, age at menarche, parity and age at first birth, height, BMI, first-degree family history of breast cancer, benign breast disease, alcohol consumption, physical activity, menopausal status, age at menopause and HRT use | |
Controls: 75,828 selected cohort members with complete statin use information | Statin use for <2 y vs none: 0.86 (0.68-1.08), statin use for 2-4 y vs none: 0.99 (0.75-1.31), and statin use for >4 y vs none: 0.93 (0.60-1.44) | ||||||
Dale et al. (61) USA | N/A | Meta-analysis | N/A | Literature database search through July 2005 of randomized clinical trials | 5 studies of breast cancer incidence representing 145 cases | N/A | |
27 of 8,943 potential articles were analyzed, representing 86,936 participants | Statin use vs none: 1.33 (0.79-2.26) | ||||||
Bonovas et al. (60) N/A | N/A | Meta-analysis | N/A | Literature database search through March 2005 of randomized clinical trials or observational studies | Statin use vs none, in fixed effects model: 1.03 (0.93-1.14) | N/A | |
16 of 683 potential articles were analyzed | Statin use vs none, in random effects model: 1.02 (0.89-1.18) | ||||||
Hwang et al. (86) USA | 1993-2004 | Cohort study | Cases: 4,383 incident cases of self-reported breast cancer confirmed by medical record and pathology review | In-person interview, medical record data | Statin use vs none: 0.91 (0.80-1.05) | Age, BMI, race, smoking, family history of breast cancer, education, hysterectomy, mammogram in the last 2 y, age at menarche, parity/age at first birth, alcohol use, percentage of calories from fat, physical activity, and NSAID use | |
Controls: 156,351 cohort members | Hydrophobic statins (Zocor, Mevacor, or Pravachol) vs none: 0.82 (0.70-0.97) | ||||||
Risk did no vary significantly by dose, duration, and HRT use at baseline | |||||||
Tumor characteristics were similar across statin users and nonusers | |||||||
Setoguchi et al. (91) USA | 1994-2003 | Registry-based study | Cohort of 31,723 adults with initiation of statin use (24,439) or glaucoma medication use (7,284) | Epidemiologic prescription database | Statin use vs glaucoma medication use: 0.99 (0.74-1.33) | Age, sex, race, Charlson comorbidity score, physician visits, total medications used, hospitalizations, prior nursing home stay, mammography, gynecologic examination, colonoscopy, fecal occult blood testing, osteoporosis drug use, arthritis, diabetes, inflammatory bowel disease, benign breast disease, HRT use, NSAID use, gastroprotective drug use, obesity, tobacco abuse | |
Cases: 268 individuals with primary invasive breast cancer | |||||||
Boudreau et al. (92) USA | 1990-2004 | Registry-based study | Cohort of 92,788 women ages 45-89 in a large health plan | Insurance plan prescription database | Statin use ever vs never: 1.07 (0.88-1.29) | Age, HRT use, diabetes, use of other lipid-lowering drugs, BMI | |
2,707 incidence invasive breast cancer cases identified through Surveillance, Epidemiology and End Results | Duration ≥5 y: 1.27 (0.89-1.81) | ||||||
Ever statin use among ER+: 1.06 (0.85-1.32) | |||||||
Duration ≥5 y among ER+: 1.24 (0.83-1.86) | |||||||
Ever statin use among ER-: 1.28 (0.78-2.08) | |||||||
Duration ≥5 y among ER-: 1.81 (0.75-4.36) | |||||||
Coogan et al. (55) USA | 1991-2005 | Hospital-based case-control study | Cases: 1,185 women with incident invasive breast cancer admitted to a participating hospital | In-person interview | Regular statin use vs never use: 1.2 (0.8-1.8) | Age, interview year, study center, BMI, alcohol consumption, race, education, tobacco use, NSAID use, HRT use, oral contraceptive use, menopausal status, parity, age at menarche, family history of breast cancer, religion | |
Controls: 2,081 women admitted to a participating hospital without cancer or disorders related to statin use | Statin use duration ≥5 y: 1.5 (0.7-3.2) |
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Peeters et al. (52) Denmark | 1991-1994 | Registry-based study | 6 cases were identified using a population-based prescription database and the Danish Cancer Registry | Epidemiologic prescription database | Use of statins vs none, standardized incidence ratio: 1.4 (0.5-3.1) | ||
1,882 patients in cohort, with 4,580 person-years of follow-up | |||||||
Jick et al. (46) Quebec | 1988-1994 | Registry-based study | Cases: 56 breast cancer cases were identified using computerized health databases of the Regie de l'Assurance-Maladie du Quebec | Computerized health record database | Ever used statin vs use of bile acid binding resins: 0.67 (0.33-1.38) | Age at index date, previous neoplasm, year of cohort entry, use of fibric acid, use of other lipid-reducing agents, and a comorbidity score | |
Controls: 560 cancer-free controls matched to cases (6,721 patients in cohort) | |||||||
Michels et al. (51) UK | 1992-1998 | Registry-based study | Cases: 224 incident invasive and in situ carcinomas from the General Practice Research Database | General practice database/electronic medical record | Current statin use vs none: 1.0 (0.6-1.6) | ||
Controls: 1,009 cancer-free matched controls | Past statin use vs none: 1.3 (0.6-2.8) | ||||||
Statin use >5 y vs none: 1.1 (0.4-3.0) | |||||||
Li et al. (80) USA | 1987-2001 | Hospital-based case-control study | Cases: 1,132 primary invasive and in situ breast cancer confirmed by pathology report | In-person interview | For all breast cancers, use of statins ≥ 3 y vs none: 2.1 (1.1-4.0) | Age, year of interview, study center, education, number of doctor visits two y before hospitalization, use of conjugated estrogens, HRT use, oral contraceptive use, religion, race, alcohol consumption, and BMI | |
Controls: 589 women with noncancer, non-statin-related conditions | For carcinoma in situ, use of statins ≥3 y vs none: 3.4 (1.5-8.0) | ||||||
For invasive breast cancer, use of statins ≥3 y vs none: 1.5 (0.7-3.1) | |||||||
Alshafie et al. (87) USA | 1992-2001 | Cohort study | 244 incident breast cancer cases confirmed by medical record and pathology report | Questionnaire and interviews | Ever used statins vs none: 0.28 (0.09-0.86) | Age and body weight | Adjustment for following variables did not appreciably effect risk estimates: HRT use, family history of breast cancer, mammography use, height, education, health status, age at menarche, age at first birth, parity, physical activity, and alcohol consumption |
7,284 cancer-free cohort members | Ever used nonstatin lipid-lowering drug vs none: 0.37 (0.14-0.99) | ||||||
Ever used any lipid-lowering drug vs none: 0.32 (0.15-0.68) | |||||||
Gonzalez-Perez et al. (45) Canada | 1989-1997 | Registry-based study | 879 incidence breast cancers identified through regional cancer registry | Computerized health record database | Ever used statins vs none: 1.09 (0.93-1.28) | None | |
Cancer-free cohort members (total cohort = 67,472) | Age ≤55 y and ever used statins vs none: 0.81 (0.53-1.24) | ||||||
Age >55 y and ever used statins vs none: 1.15 (0.97-1.37) | |||||||
Age >55 and ≥37 y HRT use vs none: 2.04 (1.20-3.46) | |||||||
Boudreau et al. (56) USA | 1997-1999 | Population-based case-control study | Cases: 975 primary, invasive cancers identified via tumor registry/Surveillance, Epidemiology and End Results | In-person interview | Ever used statins vs none: 0.9 (0.7-1.2) | Age, reference year, county of residence, use of antihypertensive medication | |
Controls: 1,007 cancer-free general population controls identified via Medicare/Medicaid lists | Current statin use >5 y vs none: 0.7 (0.4-1.0) | ||||||
Kaye and Jick (126) UK | 1990-2002 | Registry-based study | Cases: 3,224 incident cancer cases, including 698 breast cancers from the General Practice Research Database | General practice database/electronic medical record, self-administered questionnaire | Current statin use vs none: 0.9 (0.6-1.3) | Not specified for breast model, but considered BMI, smoking status, average general practice visit frequency during follow-up | |
Controls: 14,844 cancer-free matched controls | |||||||
Graaf et al. (127) The Netherlands | 1991-1998 | Registry-based study | Cases: 3,129 incident cancer cases, including 467 breast cancers from the PHARMO drug dispensing database system | Drug dispensing database linked to hospital discharge record | Ever statin use vs none: 1.07 (0.65-1.74) | Diabetes mellitus, prior hospitalizations, chronic disease score, chronic use of diuretics, ACEi, CCB, hormones, NSAID, and other lipid-lowering therapy | |
Controls: 16,976 cancer-free matched controls | |||||||
Olsen et al. (48) Denmark | 1989-2002 | Registry-based study | Cases: 22,512 incident cancer cases, including 3,141 breast cancer cases identified via Central Population Register, Epidemiologic Prescription Database, and Danish Cancer Registry | Epidemiologic prescription database | Ever statin use vs none: 1.02 (0.76-1.36) | Age, calendar period, NSAID, HRT, cardiovascular drugs | |
Controls: 334,754 men and women in general population, with 12,251 statin users | |||||||
Brueggemeier et al. (85) USA | 1994-2000 | Cohort study | Cases: 3,177 incident cases of breast cancer identified from self-report and medical record review | Self-administered questionnaire | Current statin use vs none: 0.91 (0.76-1.08) | Age, age at menarche, parity and age at first birth, height, BMI, first-degree family history of breast cancer, benign breast disease, alcohol consumption, physical activity, menopausal status, age at menopause and HRT use | |
Controls: 75,828 selected cohort members with complete statin use information | Statin use for <2 y vs none: 0.86 (0.68-1.08), statin use for 2-4 y vs none: 0.99 (0.75-1.31), and statin use for >4 y vs none: 0.93 (0.60-1.44) | ||||||
Dale et al. (61) USA | N/A | Meta-analysis | N/A | Literature database search through July 2005 of randomized clinical trials | 5 studies of breast cancer incidence representing 145 cases | N/A | |
27 of 8,943 potential articles were analyzed, representing 86,936 participants | Statin use vs none: 1.33 (0.79-2.26) | ||||||
Bonovas et al. (60) N/A | N/A | Meta-analysis | N/A | Literature database search through March 2005 of randomized clinical trials or observational studies | Statin use vs none, in fixed effects model: 1.03 (0.93-1.14) | N/A | |
16 of 683 potential articles were analyzed | Statin use vs none, in random effects model: 1.02 (0.89-1.18) | ||||||
Hwang et al. (86) USA | 1993-2004 | Cohort study | Cases: 4,383 incident cases of self-reported breast cancer confirmed by medical record and pathology review | In-person interview, medical record data | Statin use vs none: 0.91 (0.80-1.05) | Age, BMI, race, smoking, family history of breast cancer, education, hysterectomy, mammogram in the last 2 y, age at menarche, parity/age at first birth, alcohol use, percentage of calories from fat, physical activity, and NSAID use | |
Controls: 156,351 cohort members | Hydrophobic statins (Zocor, Mevacor, or Pravachol) vs none: 0.82 (0.70-0.97) | ||||||
Risk did no vary significantly by dose, duration, and HRT use at baseline | |||||||
Tumor characteristics were similar across statin users and nonusers | |||||||
Setoguchi et al. (91) USA | 1994-2003 | Registry-based study | Cohort of 31,723 adults with initiation of statin use (24,439) or glaucoma medication use (7,284) | Epidemiologic prescription database | Statin use vs glaucoma medication use: 0.99 (0.74-1.33) | Age, sex, race, Charlson comorbidity score, physician visits, total medications used, hospitalizations, prior nursing home stay, mammography, gynecologic examination, colonoscopy, fecal occult blood testing, osteoporosis drug use, arthritis, diabetes, inflammatory bowel disease, benign breast disease, HRT use, NSAID use, gastroprotective drug use, obesity, tobacco abuse | |
Cases: 268 individuals with primary invasive breast cancer | |||||||
Boudreau et al. (92) USA | 1990-2004 | Registry-based study | Cohort of 92,788 women ages 45-89 in a large health plan | Insurance plan prescription database | Statin use ever vs never: 1.07 (0.88-1.29) | Age, HRT use, diabetes, use of other lipid-lowering drugs, BMI | |
2,707 incidence invasive breast cancer cases identified through Surveillance, Epidemiology and End Results | Duration ≥5 y: 1.27 (0.89-1.81) | ||||||
Ever statin use among ER+: 1.06 (0.85-1.32) | |||||||
Duration ≥5 y among ER+: 1.24 (0.83-1.86) | |||||||
Ever statin use among ER-: 1.28 (0.78-2.08) | |||||||
Duration ≥5 y among ER-: 1.81 (0.75-4.36) | |||||||
Coogan et al. (55) USA | 1991-2005 | Hospital-based case-control study | Cases: 1,185 women with incident invasive breast cancer admitted to a participating hospital | In-person interview | Regular statin use vs never use: 1.2 (0.8-1.8) | Age, interview year, study center, BMI, alcohol consumption, race, education, tobacco use, NSAID use, HRT use, oral contraceptive use, menopausal status, parity, age at menarche, family history of breast cancer, religion | |
Controls: 2,081 women admitted to a participating hospital without cancer or disorders related to statin use | Statin use duration ≥5 y: 1.5 (0.7-3.2) |
Antihypertensive Medication Use and Risk of Breast Cancer
Biological Mechanisms
Research into the biological mechanisms by which antihypertensive agents may affect carcinogenesis has focused on calcium channel blockers (CCB) and angiotensin II–converting enzyme inhibitors (ACEi). Pahor et al. have suggested that CCB could play a role in increased cancer risk (62) due to inhibition of apoptosis resulting from diminished intracellular calcium ion concentrations (63-65). However, as reviewed by Mason et al. (66), the role of calcium ions in apoptosis has been shown to be inconsistent, with intracellular calcium levels yielding both increased and decreased apoptosis across a range of cell types. Additionally, research has shown that CCB may actually inhibit carcinogenesis by limiting cell proliferation in breast cell lines (67, 68), making it difficult to draw firm conclusions about their ultimate effect on cancer risk.
ACEi have been suggested to offer a potential protective effect against cancer risk through the inhibition of angiogenesis. More specifically, ACEi target the action of angiotensin II, part of the rennin-angiotensin system involved with renal blood flow, fluid homeostasis, and blood pressure control (69). Angiotensin II has also been shown to promote neovascularization (70), a necessary process for tumor development. Early studies showed that angiogenesis and tumor growth were slowed following administration of ACEi in preclinical studies (71, 72). Later, Yoshiji et al. (73) hypothesized that the inhibition of angiotensin II interferes with the action of vascular endothelial growth factor, a key enzyme in the angiogenesis process. Although cell proliferation has not shown to be directly effected (74), use of ACEi alone or in combination with other agents decreased vascular endothelial growth factor concentrations and angiogenesis (75-77) and reduced blood vessel and formation around tumors (74).
Summary of Existing Evidence
An increasing number of studies have focused on the potential role of antihypertensive drug use in breast cancer development (Table 4). These studies have largely focused on CCB, β-blockers, and ACEi, and we will restrict our discussion to these widely studied drugs. As with many pharmacoepidemiologic efforts, most of these prior studies were registry based such as general practice database or electronic medical records and used data from health-care plan records or prescription plan. The limitations of this approach are outlined above. Nevertheless, results from these studies do not indicate that ever or prolonged use of CCB, β-blockers, or ACEi was related to elevated breast cancer risk (45-49, 78, 79). Similarly, results from a large hospital-based case-control study (50), the Nurses' Health Study cohort (51), and a Dutch cohort study (52) do not suggest that these drugs are related to breast cancer risk. In contrast, findings from a smaller cohort study (53) have linked ever use of CCB to a significant increase in risk (OR, 2.57; 95% CI, 1.47-4.49). No risk elevations were observed for use of β-blockers and ACEi. Li et al. (80), in a large population-based case-control study, observed a significant increase in risk for prolonged use (≥15 years) of β-blockers (OR, 2.1; 95% CI, 1.2-3.7) but no associations with long-term use of CCB and ACEi. Finally, Largent et al. (81) recently reported results from another population-based case-control study. Results indicated that ever (OR, 1.79; 95% CI, 1.07-3.01) and prolonged (OR, 3.50; 95% CI, 1.64-7.50) use of diuretics was associated with excess risk. No such risk elevations were observed for nondiuretic antihypertensive medications.
Epidemiologic studies of the role of antihypertensive drug use in breast cancer development
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Schreinemachers and Everson (110) UK | 1995 | Registry-based study | Cases: 80 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record, self-administered questionnaire | Ever used CCB vs β-blocker users: 1.32 (0.72-2.41) | Smoking, BMI, change of medication, duration of hypertension, diuretic use | |
Database | |||||||
Controls: 1,750 total cancer-free controls frequency matched by age and practice location | |||||||
Egan et al. (117) USA | 1989-1996 | Cohort study | 75 primary invasive cases confirmed by medical record abstraction | Standardized questionnaire | Ever used CCB vs none: 2.57 (1.47-4.49) | Age, race, parity, age at menopause, self-reported diabetes | |
3,123 cancer-free cohort members | Ever used β-blocker vs none: 1.14 (0.58-2.25) | ||||||
Ever used ACEi vs none: 0.93 (0.37-2.34) | |||||||
Ever used any diuretic vs none: 1.38 (0.83-2.29) | |||||||
Ever used any vasodilator vs none: 0.30 (0.07-1.21) | |||||||
Olsen et al. (48) Denmark | 1991-1993 | Registry-based study | 32 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used CCB: standardized incidence ratio: 0.8 (0.5-1.1) | None | |
17,911 patients in cohort, including men and women (32,540 person-years of follow-up) | |||||||
Peeters et al. (52) The Netherlands | 1974-1985 | Cohort | Cohort of 11,075 women ages 50-65 y enrolled in a breast cancer screening project | Self-administered questionnaires | No increased in mortality from breast cancer for use of any antihypertensive drug (data not shown) | ||
114 cases of breast cancer identified | |||||||
Jacobs et al. (115) USA | 1988-1994 | Cohort study | 355 self-reported cancers confirmed by medical record abstraction | Self-administered questionnaire | Use of CCB vs none: 1.07 (0.78-1.48) | Age, multiple drug use, self-reported weight, height, smoking status and mean number of cigarettes smoked per day among women who smoked in 1988, alcohol intake in 1988, physical activity, menopausal status in 1988, postmenopausal HRT use, cholesterol level, systolic and diastolic blood pressure in 1988, aspirin intake, diabetes, history of stroke, myocardial infarction, CABG/PTCA, angina, hypertension in or before 1988, family history of breast cancer, history of benign breast disease, age at menarche, parity, age at first birth, age at menopause | |
18,635 cohort members in analysis (107,256 person-years of follow-up) | |||||||
Ready et al. (114) USA | 1983-1996 | Hospital-based case-control study | Cases: 2,893 primary breast cancer cases confirmed from discharge summaries and pathology reports | In-person interview | Use of CCB ≥5 y vs none: 1.1 (0.7-1.8) | Age, study center, interview year, BMI, annual visits to physician before diagnosis, race, years of education, breast cancer in mother or sister, benign breast disease, age at menarche, age at first birth, parity, age at menopause, alcohol consumption, duration of oral contraceptive use, duration of HRT use | |
Controls: 6,492 controls admitted for nonmalignant conditions | Use of β-blockers ≥5 y vs none: 1.1 (0.9-1.5) | ||||||
Use of ACEi ≥5 y vs none: 1.2 (0.7-2.2) | |||||||
Harris et al. (113) Denmark | 1989-1995 | Registry-based study | 84 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used CCB: standardized incidence ratio: 0.97 (0.77-1.20) | None | |
23,167 cohort members, including men and women (73,193 person-years follow-up) | |||||||
Johnson et al. (111) UK | 1992-1997 | Registry-based study | Cases: 3,706 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | ACEi use ≥5 y vs none: 1.0 (0.7-1.5) | BMI, smoking status | Adjustment for following variables did not appreciably effect risk estimates: alcoholism, hysterectomy, and breast lumps |
Database | CCB use ≥5 y vs none: 0.9 (0.7-1.2) | ||||||
Controls: 14,155 cancer-free controls from cohort matched 4:1 to cases on age, physician practice, index date, number of years of medical history record in database | β-blocker use ≥5 y vs none: 1.0 (0.8-1.2) | ||||||
Gallicchio et al. (107) Denmark | 1989-1995 | Registry-based study | 83 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used ACEi (with previous use of CCB and/or β-blocker): standardized incidence ratio: 1.1 (0.9-1.3) | None | |
17,897 cohort members, including men and women (66,827 person-years follow-up) | Exclusive ACEi use: standardized incidence ratio: 1.1 (0.8-1.5) | ||||||
Shen et al. (109) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | Used diuretics > 3 y vs none: 1.1 (0.9-1.2) | Age, calendar year, hypertension, BMI, alcohol intake, smoking status, HRT use, and prior breast lump and/or breast biopsy | |
Database | Used β-blockers > 3 y vs none: 1.1 (0.9-1.2) | ||||||
Controls: 20,000 cancer-free controls from cohort matched to cases on age and calendar year (study cohort = 734,899 women) | Used ACEi > 3 y vs none: 0.9 (0.7-1.2) | ||||||
Used CCB >3 y vs none: 1.0 (0.8-1.2) | |||||||
Used α-blockers >3 y vs none: 0.2 (0.2-1.3) | |||||||
Friis et al. (82) USA | 1997-1999 | Population-based case-control study | Cases: 975 cases of invasive breast cancer identified via Cancer Surveillance System, a population-based cancer registry | In person interview | Used CCB for 15 y vs none: 0.6 (0.3-1.3) | Reference age | Adjustment for following variables did not appreciably effect risk estimates: race, income, marital status, education, age at menarche, parity, age at first birth, type of menopause, age at menopause, duration of oral contraceptive use, HRT use, first-degree family history of breast cancer, smoking status, average daily alcohol intake, and BMI |
Controls: 1,007 cancer-free controls identified from list of Medicare/Medicaid recipients, selected for similar age | Used β-blockers for ≥15 y vs none: 2.1 (1.2-3.7) | ||||||
Used ACEi for ≥15 y vs none: 0.8 (0.4-1.6) | |||||||
Used diuretics for ≥15 y vs none: 1.2 (0.8-1.6) | |||||||
Moorman et al. (108) Denmark | 1990-2002 | Registry-based study | 264 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used any antihypertensive vs never: 0.95 (0.81-1.10) | Age, calendar period, HRT use, NSAID use, parity, and age at first birth | |
49,950 women in total cohort (19,284 statin users contributing 109,985 person-years of follow-up) | Ever used ace inhibitor vs none: 0.99 (0.75-1.31) | ||||||
Ever used angiotensin II agonist vs none: 1.01 (0.67-1.51) | |||||||
Ever used β-blockers vs none: 0.98 (0.79-1.22) | |||||||
Ever used CCB vs none: 0.80 (0.59-1.09) | |||||||
Ever used diuretic vs none: 0.95 (0.8-1.12) | |||||||
Risk estimates not significantly effected by number of prescriptions, years of follow-up, type of diuretic, or type of calcium antagonist | |||||||
Largent et al. (81) USA | 1994-1995 | Population-based case-control study | Cases: 523 women age 50-75 y with incident breast cancer | Self-administered questionnaire | Diuretic use ever vs never: 1.79 (1.07-3.01) | Age, BMI, diabetes, smoking, alcohol use, menopausal status, family history of breast or ovarian cancer, age at first pregnancy, education | |
Controls: 131 women ages 50-75 y old identified through random-digit dialing, matched to cases on age | Diuretic duration ≥6 y vs never: 3.50 (1.64-7.50) | ||||||
Use of nondiuretic antihypertensive drug ever vs never: 1.18 (0.69-2.03) | |||||||
Nondiuretic antihypertensive duration ≥6 y vs never: 1.24 (0.62-2.50) |
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Schreinemachers and Everson (110) UK | 1995 | Registry-based study | Cases: 80 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record, self-administered questionnaire | Ever used CCB vs β-blocker users: 1.32 (0.72-2.41) | Smoking, BMI, change of medication, duration of hypertension, diuretic use | |
Database | |||||||
Controls: 1,750 total cancer-free controls frequency matched by age and practice location | |||||||
Egan et al. (117) USA | 1989-1996 | Cohort study | 75 primary invasive cases confirmed by medical record abstraction | Standardized questionnaire | Ever used CCB vs none: 2.57 (1.47-4.49) | Age, race, parity, age at menopause, self-reported diabetes | |
3,123 cancer-free cohort members | Ever used β-blocker vs none: 1.14 (0.58-2.25) | ||||||
Ever used ACEi vs none: 0.93 (0.37-2.34) | |||||||
Ever used any diuretic vs none: 1.38 (0.83-2.29) | |||||||
Ever used any vasodilator vs none: 0.30 (0.07-1.21) | |||||||
Olsen et al. (48) Denmark | 1991-1993 | Registry-based study | 32 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used CCB: standardized incidence ratio: 0.8 (0.5-1.1) | None | |
17,911 patients in cohort, including men and women (32,540 person-years of follow-up) | |||||||
Peeters et al. (52) The Netherlands | 1974-1985 | Cohort | Cohort of 11,075 women ages 50-65 y enrolled in a breast cancer screening project | Self-administered questionnaires | No increased in mortality from breast cancer for use of any antihypertensive drug (data not shown) | ||
114 cases of breast cancer identified | |||||||
Jacobs et al. (115) USA | 1988-1994 | Cohort study | 355 self-reported cancers confirmed by medical record abstraction | Self-administered questionnaire | Use of CCB vs none: 1.07 (0.78-1.48) | Age, multiple drug use, self-reported weight, height, smoking status and mean number of cigarettes smoked per day among women who smoked in 1988, alcohol intake in 1988, physical activity, menopausal status in 1988, postmenopausal HRT use, cholesterol level, systolic and diastolic blood pressure in 1988, aspirin intake, diabetes, history of stroke, myocardial infarction, CABG/PTCA, angina, hypertension in or before 1988, family history of breast cancer, history of benign breast disease, age at menarche, parity, age at first birth, age at menopause | |
18,635 cohort members in analysis (107,256 person-years of follow-up) | |||||||
Ready et al. (114) USA | 1983-1996 | Hospital-based case-control study | Cases: 2,893 primary breast cancer cases confirmed from discharge summaries and pathology reports | In-person interview | Use of CCB ≥5 y vs none: 1.1 (0.7-1.8) | Age, study center, interview year, BMI, annual visits to physician before diagnosis, race, years of education, breast cancer in mother or sister, benign breast disease, age at menarche, age at first birth, parity, age at menopause, alcohol consumption, duration of oral contraceptive use, duration of HRT use | |
Controls: 6,492 controls admitted for nonmalignant conditions | Use of β-blockers ≥5 y vs none: 1.1 (0.9-1.5) | ||||||
Use of ACEi ≥5 y vs none: 1.2 (0.7-2.2) | |||||||
Harris et al. (113) Denmark | 1989-1995 | Registry-based study | 84 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used CCB: standardized incidence ratio: 0.97 (0.77-1.20) | None | |
23,167 cohort members, including men and women (73,193 person-years follow-up) | |||||||
Johnson et al. (111) UK | 1992-1997 | Registry-based study | Cases: 3,706 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | ACEi use ≥5 y vs none: 1.0 (0.7-1.5) | BMI, smoking status | Adjustment for following variables did not appreciably effect risk estimates: alcoholism, hysterectomy, and breast lumps |
Database | CCB use ≥5 y vs none: 0.9 (0.7-1.2) | ||||||
Controls: 14,155 cancer-free controls from cohort matched 4:1 to cases on age, physician practice, index date, number of years of medical history record in database | β-blocker use ≥5 y vs none: 1.0 (0.8-1.2) | ||||||
Gallicchio et al. (107) Denmark | 1989-1995 | Registry-based study | 83 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used ACEi (with previous use of CCB and/or β-blocker): standardized incidence ratio: 1.1 (0.9-1.3) | None | |
17,897 cohort members, including men and women (66,827 person-years follow-up) | Exclusive ACEi use: standardized incidence ratio: 1.1 (0.8-1.5) | ||||||
Shen et al. (109) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | Used diuretics > 3 y vs none: 1.1 (0.9-1.2) | Age, calendar year, hypertension, BMI, alcohol intake, smoking status, HRT use, and prior breast lump and/or breast biopsy | |
Database | Used β-blockers > 3 y vs none: 1.1 (0.9-1.2) | ||||||
Controls: 20,000 cancer-free controls from cohort matched to cases on age and calendar year (study cohort = 734,899 women) | Used ACEi > 3 y vs none: 0.9 (0.7-1.2) | ||||||
Used CCB >3 y vs none: 1.0 (0.8-1.2) | |||||||
Used α-blockers >3 y vs none: 0.2 (0.2-1.3) | |||||||
Friis et al. (82) USA | 1997-1999 | Population-based case-control study | Cases: 975 cases of invasive breast cancer identified via Cancer Surveillance System, a population-based cancer registry | In person interview | Used CCB for 15 y vs none: 0.6 (0.3-1.3) | Reference age | Adjustment for following variables did not appreciably effect risk estimates: race, income, marital status, education, age at menarche, parity, age at first birth, type of menopause, age at menopause, duration of oral contraceptive use, HRT use, first-degree family history of breast cancer, smoking status, average daily alcohol intake, and BMI |
Controls: 1,007 cancer-free controls identified from list of Medicare/Medicaid recipients, selected for similar age | Used β-blockers for ≥15 y vs none: 2.1 (1.2-3.7) | ||||||
Used ACEi for ≥15 y vs none: 0.8 (0.4-1.6) | |||||||
Used diuretics for ≥15 y vs none: 1.2 (0.8-1.6) | |||||||
Moorman et al. (108) Denmark | 1990-2002 | Registry-based study | 264 primary invasive cases identified by population-based Epidemiologic Prescription Database and confirmed via Danish Cancer Registry | Epidemiologic prescription database | Ever used any antihypertensive vs never: 0.95 (0.81-1.10) | Age, calendar period, HRT use, NSAID use, parity, and age at first birth | |
49,950 women in total cohort (19,284 statin users contributing 109,985 person-years of follow-up) | Ever used ace inhibitor vs none: 0.99 (0.75-1.31) | ||||||
Ever used angiotensin II agonist vs none: 1.01 (0.67-1.51) | |||||||
Ever used β-blockers vs none: 0.98 (0.79-1.22) | |||||||
Ever used CCB vs none: 0.80 (0.59-1.09) | |||||||
Ever used diuretic vs none: 0.95 (0.8-1.12) | |||||||
Risk estimates not significantly effected by number of prescriptions, years of follow-up, type of diuretic, or type of calcium antagonist | |||||||
Largent et al. (81) USA | 1994-1995 | Population-based case-control study | Cases: 523 women age 50-75 y with incident breast cancer | Self-administered questionnaire | Diuretic use ever vs never: 1.79 (1.07-3.01) | Age, BMI, diabetes, smoking, alcohol use, menopausal status, family history of breast or ovarian cancer, age at first pregnancy, education | |
Controls: 131 women ages 50-75 y old identified through random-digit dialing, matched to cases on age | Diuretic duration ≥6 y vs never: 3.50 (1.64-7.50) | ||||||
Use of nondiuretic antihypertensive drug ever vs never: 1.18 (0.69-2.03) | |||||||
Nondiuretic antihypertensive duration ≥6 y vs never: 1.24 (0.62-2.50) |
Although most studies on this topic generated null findings, the majority of these investigations could only crudely classify participants as ever or never users of these drugs. Further, one study with more sophisticated exposure assessment showed an association between prolonged use of β-blockers (82). Thus, future studies employing solid epidemiologic designs and sophisticated exposure assessment might be needed to definitively rule out the role of antihypertensive medication use in breast cancer development.
NSAID Use and Breast Cancer Risk
Biological Mechanism
NSAIDs, including aspirin, ibuprofen, and naproxen, appear to exert an anticancer effect through inhibition of the COX enzyme system. COX-2, in particular, promotes the synthesis of prostaglandins, such as prostaglandin E2, thought to play an etiologic role in tissue generation and tumorigenesis. COX-2-derived prostaglandin E2 may stimulate estrogen biosynthesis in breast tissue (83). Additionally, COX-2 has been found to be overexpressed in human breast tumors in multiple studies (84-86). The observation that COX-2 expression is correlated with aromatase expression in breast cancer allows one to formulate the hypothesis that COX-2 increases estrogen production via up-regulation of aromatase expression. Preclinical research has shown that the administration of NSAID inhibits production of COX enzymes with resulting reduction in mammary carcinogenesis (87-89). Moreover, NSAIDs have been suggested to reduce neovascularization and promote apoptosis (63, 90). Some NSAIDs that do not affect the COX system have been shown to induce cell cycle arrest and apoptosis in breast cancer cell lines (64). Taken together, multiple lines of research into the biological mechanisms by which NSAIDs affect cancer risk point to a potentially valid agent in chemoprevention.
Summary of Existing Evidence
A large and diverse body of literature exists on the potential chemopreventive effect of NSAID use on breast cancer development (Table 5). Exposure assessment, however, differs widely across studies, including the definition of regular and prolonged use. Nevertheless, results from most studies have been remarkably consistent. Three registry-based studies (91-93) showed significant risk reductions for prolonged NSAID use. Several hospital-based (65, 94-97) and population-based (98-102) studies have generated statistically significant risk reductions for regular and prolonged aspirin use, except for a recent one (103). Less consistent evidence exists for ibuprofen use, which was associated with decreased risk in two investigations (104, 133) but not in others (131, 136, 138). Such discrepancy might not be surprising, given that ibuprofen is still a relatively new drug, and to date, few people have had significant exposures to this agent. Findings from the Women's Health Initiative observational study indicated that prolonged use (≥10 years) of any NSAID or aspirin was associated with statistically significant risk reductions (RR, 0.72; 95% CI, 0.56-0.91 and RR, 0.79; 95% CI, 0.60-1.03, respectively; ref. 105). Similarly, findings from the CLUE cohort in Washington county (106) point to a chemoprotective effect of aspirin use in breast cancer etiology (RR, 0.46; 95% CI, 0.22-0.98), but results were not influenced by hormone receptor status or COX-2 genetic polymorphisms (107). Other studies have also attempted to assess the effect of the COX-2 gene on the association between NSAID use and breast cancer risk, but results have been inconsistent (108, 109). Further support for a chemopreventive role of aspirin comes from the National Health and Nutrition Examination Survey I cohort (110) and Iowa Women's cohort (111) where current or prolonged (≥6 years) use were associated with significant risk decreases (RR, 0.70; 95% CI, 0.56-0.96 and RR, 0.71; 95% CI, 0.58-0.87, respectively). In the Iowa Women's cohort, these risk reductions were still apparent in subsequent analyses based on more breast cancer patients (112). These findings are similar to those of a smaller cohort from Ohio (113), where frequent NSAID use was associated with a significant risk reduction (RR, 0.57; 95% CI, 0.44-0.74). Recently, Ready et al. (114) found significant risk reduction for frequent and long-term use of low-dose aspirin (≥4 days/wk over 10 years) in the Vitamins and Lifestyle cohort (RR, 0.65; 95% CI, 0.43-0.91).
Epidemiologic studies of the role of NSAID use in breast cancer development
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Schreinemachers and Everson (110) USA | 1971-1975 | Cohort | 1,257 cases identified via the National Health and Examination Survey I | In-person and telephone interviews, hospital and nursing home records | Incident risk ratio for all sites combined for aspirin users vs nonaspirin users (30 d before interview) | Gender, age | Adjustment for following variables did not appreciably effect incident risk ratio: race, education, smoking, alcohol |
11,411 cancer-free cohort members | All sites combined: 0.83 (0.74-0.93), lung cancer: 0.68 (0.49-0.94), breast cancer in women: 0.70 (0.50-0.96), and colorectal cancer in younger men: 0.35 (0.17-0.73) | ||||||
Harris et al. (95) USA | 1988-1992 | Hospital-based case-control study | Cases: 744 patients with newly diagnosed breast cancer identified by collaborating hospitals in northeastern United States | In-person interview | 1-4 y NSAID use vs none: 1.09 (0.8-1.5), ≥5 y NSAID use vs none: 0.63 (0.5-0.9) | Age, menopausal status, parity, family history of breast cancer, BMI | |
Controls: 767 patients without cancer diagnoses frequency matched to cases | |||||||
Egan et al. (117) USA | 1980-1992 | Cohort | 2,414 cases of invasive breast cancer (2,303 confirmed with medical records and 111 cases identified by questionnaire response) | Self-administered questionnaire, medical record review | Regular aspirin use from 1980 to 1988 vs no regular use: 1.01 (0.80-1.27) | Age at menarche, age at menopause, BMI, alcohol, family history of breast cancer, history of benign breast disease, multivitamin use | Authors concluded that regular aspirin use does not reduce breast cancer risk |
Heavy use from 1980 to 1988 vs no regular use: 1.09 (0.75-1.60) | |||||||
≥20 y regular use vs no regular use: 1.00 (0.71-1.41) | |||||||
Harris et al. (99) USA | Not noted | Population-based case-control study | Cases: 511 newly diagnosed breast cancer confirmed by pathology report | In-person interview | Regular NSAID use vs no regular use: 0.66 (0.52-0.83) | Age, parity, menopausal status, family history | |
Controls: 1,534 cancer-free women from central OH frequency matched by race and age | Regular aspirin use vs no regular use: 0.69 (0.46-0.99) | ||||||
Regular ibuprofen use vs no regular use: 0.57 (0.36-0.91) | |||||||
≥7 per week, ≥5 y NSAID use vs no regular use: 0.60 (0.40-0.91) | |||||||
Harris et al. (113) USA | 1991-1993 | Cohort | 393 breast cancers have been detected | Self-administered questionnaire | 1-3 NSAID pills per week vs <1: 0.64 (0.50-0.82), ≥4 vs <1: 0.57 (0.44-0.74) | None | Adjustment for the following variables did not appreciably effect risk estimates: age, education, parity, menopausal status, and family history of breast cancer |
32,505 women enrolled in the mammography screening program of The Ohio State University Comprehensive Cancer Center (4.7 y average follow-up) | |||||||
Coogan et al. (65) USA | 1976-1996 | Hospital-based case-control study | Cases: 6,558 women with a first occurrence of primary breast cancer diagnosed within the previous year, confirmed by path report, and no concurrent or previous cancer | In-person interview | For cancer controls: regular use within 1 y of admission only vs never use: 0.6 (0.4-1.0); regular use begun ≥1 y before admission vs never use: 0.8 (0.7-1.0) | Age, study center, interview year, years of education, history of benign breast disease, number of doctor visits 2 y before admission, duration of oral contraceptive use, duration of use of female hormone supplements | Adjustment for the following variables did not appreciably effect risk estimates: age at menarche, age at menopause, age at first birth, parity, race, alcohol consumption, religion, breast cancer in mother or sister, practice of breast self-examination, BMI |
Controls: 3,296 patients with other cancers not associated with NSAID use, 2,925 noncancer patients | For noncancer controls: regular use within 1 y of admission only vs never use: 0.5 (0.3-0.8), regular use begun ≥1 y before admission vs never use: 0.7 (0.6-0.9) | ||||||
Sharpe et al. (93) Canada | 1981-1995 | Registry-based study | Cases: 5,882 women diagnosed with histologically proven invasive breast cancer | Saskatchewan Prescription Drug Plan database | NSAID exposure 2-5 y before diagnosis: average daily dose >0.3 vs ADD = 0: 0.76 (0.63-0.92) | Sampling fractions, age, exposure during other periods, total duration of lactation, BMI after menopause | For periods <2 and >5 y, there was no significant reduction in risk |
Controls: 23,517 controls frequency matched on age and sampling time | Just cases, exposure 2-5 y before diagnosis: 0 < ADD ≤ 0.1 vs ADD = 0: 0.52 (0.37-0.73), 0.1 < ADD ≤ 0.3 vs ADD = 0: 0.53 (0.30-0.92), and ADD > 0.3 vs ADD = 0: 0.49 (0.24-0.99) | ||||||
Khuder and Mutgi (120) N/A | N/A | Meta-analysis | N/A | Literature database search through 2000 | NSAID use vs none | N/A | The numbers given in Table 3 are different than the numbers presented in the body of the article |
14 articles were analyzed | 0.82 (0.75-0.89) in all studies, 0.78 (0.62-0.99) in 6 cohort studies, and 0.87 (0.84-0.91) in 8 case-control studies | ||||||
Cotterchio et al. (98) Canada | 1996-1998 | Population-based case-control study | Cases: 3,133 random women diagnosed with a first primary cancer of the breast, 25-74 y identified via Ontario Cancer Registry | Self-report questionnaire | Any regular NSAID use vs never: 0.76 (0.66-0.88) | Age, history of arthritis, benign breast disease | Confounders evaluated include HRT, oral contraceptive use, alcohol, smoking, weight, BMI, physical activity, history of arthritis, reproductive history, education, marital status, previous breast cysts, family history of breast cancer, other medication use, dietary fat intake |
Controls: 3,062 age-matched random sample of the female population of Ontario | ≥9 y NSAID use vs never: 0.68 (0.54-0.86) | ||||||
≤1 y since last NSAID use vs never: 0.64 (0.54-0.77) | |||||||
Age at first use ≥50 vs never: 0.76 (0.61-0.93) | |||||||
Meier et al. (128) UK | Registry-based study | Cases: 3,706 women with incident breast cancer | Medical history computer record | 20-29 acetaminophen prescriptions vs none: 0.7 (0.6-0.9) | Smoking status, BMI | Adjustment for following variables did not appreciably effect risk estimates: prior hysterectomy, prior oophorectomy, prior history of benign breast lumps, longer-term exposure to postmenopausal estrogens | |
Controls: 14,155 age, years of medical history in the computer record, general practice attended, and calendar time matched controls | ≥30 acetaminophen prescriptions vs none: 0.8 (0.7-1.0) | Study contains no information about over-the-counter NSAID | |||||
No statistically significant difference found between number of NSAID prescriptions | |||||||
Johnson et al. (111) USA | 1986-1997 | Cohort | 938 cases identified from the Iowa Women's Health Study | Self-administered questionnaire | Aspirin or NSAID use vs none: 0.80 (0.67-0.95) | Age, BMI, estrogen use, family history of breast cancer, benign breast disease, multivitamin use, category of NSAID use, mammography, waist-to-hip ratio | |
27,616 total cohort members | Aspirin use vs none: 0.82 (0.71-0.95) | ||||||
NSAID use vs none: 0.98 (0.85-1.14) | |||||||
≥6 aspirin use per week vs none: 0.71 (0.58-0.87) | |||||||
In situ disease: ≥6 aspirin use vs none: 0.52 (0.30, 0.90) | |||||||
Regional or distant disease: ≥6 aspirin use vs none: 0.50 (0.29-0.88) | |||||||
Harris et al. (105) USA | Not noted | Cohort study | 1,392 self-reported incident cases confirmed by medical record review | Self-administered questionnaire | Any NSAID use for ≥10 vs <1 y use: 0.72 (0.56-0.91) | Age, ethnicity, education, BMI, HRT use, family history of breast cancer, parity at age <30 y, and episodes of weekly exercise | Additional analyses stratified by BMI, HRT use, family history of breast cancer, parous at age <30 y, and episodes per week of moderate/strenuous exercise did not vary appreciably from full sample |
80,741 women in total cohort (43-mo average follow-up) | Aspirin use for ≥10 vs <1 y use: 0.79 (0.60-1.03) | ||||||
Ibuprofen use for ≥10 vs <1 y use: 0.51 (0.28-0.96) | |||||||
Moorman et al. (100) USA | 1996-2000 | Population-based case-control study | Cases: 930 cases of invasive breast cancer identified via North Carolina Central Cancer Registry | In-person interview | Any NSAID use vs none: 0.4 (0.3-0.6) | Age, race, age at menarche, age at first full-term pregnancy, breastfeeding history, menopausal status, family history, oral contraceptive use, HRT use, education, BMI, waist-to-hip ratio, smoking status, and offset term | |
Controls: 754 controls selected from DMV and Health Care Financing Administration, frequency matched to cases on age and ethnicity | Occasional NSAID use vs none: 0.5 (0.3-0.7) | ||||||
Regular NSAID use ≥3 y vs none: 0.3 (0.2-0.5) | |||||||
Gonzalez-Perez et al. (122) N/A | N/A | Meta-analysis | N/A | Literature database search from 1966 to 2002 for cohort or case-control studies | NSAID use vs none: 0.77 (0.66-0.88) | N/A | |
15 of 47 studies reporting outcomes for breast cancer were analyzed | Aspirin use vs none: 0.77 (0.69-0.86) | ||||||
Garcia Rodriguez and Gonzalez-Perez (104) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | Aspirin use ≥4 y vs none: 0.86 (0.61-1.19) | Age, calendar year, BMI, alcohol intake, smoking status, HRT use, prior benign breast disease, and remaining NSAID | |
Database | Nonaspirin NSAID (ibuprofen) use ≥4 y vs none: 0.94 (0.74-1.21) | ||||||
Controls: 20,000 cancer-free controls from cohort matched to cases on age and calendar year (study cohort = 734,899 women) | Acetaminophen use ≥4 y vs none: 0.77 (0.64-0.94) | ||||||
Terry et al. (102) USA | 1996-1997 | Population-based case-control study | Cases: 1,508 invasive or in situ breast cancer cases confirmed by medical record review | In-person interviews, medical records | Aspirin use ≥7 times/wk for ≥5 y vs none: 0.77 (0.57-1.04) | Age at diagnosis, migraine headache, BMI | |
Controls: 1,556 controls selected though random-digit dialing methods and Health Care Financing Administration lists, frequency matched to cases in 5-y age intervals | Ibuprofen use ≥3 times/wk for ≥5 y vs none: 1.09 (0.70-1.70) | ||||||
Ever used aspirin and hormone receptor positive vs none: 0.74 (0.60-0.93) | |||||||
Ever used aspirin and hormone receptor negative vs none: 0.97 (0.67-1.40) | |||||||
Harris et al. (160) N/A | N/A | Meta-analysis | N/A | Literature database search through 1970-2003 | All studies NSAID vs no use: RR, 0.61 (0.50-0.75) | N/A | |
17 studies | |||||||
Jacobs et al. (115) USA | 1992-2001 | Cohort study | 3008 incident cases identified via self-report and confirmed via medical record or state cancer registries | Self-administered questionnaire | ≥60 tablets of any NSAID per month vs none: 1.07 (0.96-1.21) | Age, race, education, family history of breast cancer, personal history of breast cysts, history of mammography, age at menarche, duration of oral contraceptive use, parity, age at menopause, HRT use, weight change, BMI, alcohol consumption | |
97,786 women in total cohort | ≥60 aspiring tablets per month vs none: 1.01 (0.84-1.20) | ||||||
≥60 ibuprofen tablets per month vs none: 1.06 (0.89-1.26) | |||||||
≥5 y current regular NSAID use vs none: 1.05 ((0.88-1.26) | |||||||
≥5 y current regular aspirin use vs none: 0.88 (0.69-1.12) | |||||||
≥5 y current regular ibuprofen use vs none: 1.29 (0.92-1.82) | |||||||
Zhang et al. (97) USA | 1976-2002 | Hospital-based case-control study | Cases: 7,006 primary breast cancer cases confirmed from discharge summaries and pathology reports | In-person interview | ≥20 y regular aspirin use vs none: 0.59 (0.25-1.36) | Age, year of interview, study center, race, year of education, benign breast disease, number of physician visits 2 y before hospitalization, duration of HRT use, duration of oral contraceptive use, age at menarche, age at menopause, age at first birth, parity, alcohol consumption, family history of breast cancer, practice of breast self-exam, and BMI | |
Controls: 3,622 controls admitted for nonmalignant conditions | ≥5 y regular ibuprofen use v none: 0.78 (0.29-2.08) | ||||||
Regular use of aspirin and hormone receptor positive vs none: 0.74 (0.44-1.26) | |||||||
Regular use of aspirin and hormone receptor negative vs none: 0.94 (0.45-1.96) | |||||||
Regular use any NSAID and premenopausal vs none: 0.62 (0.41-0.94) | |||||||
Regular use any NSAID and postmenopausal vs none: 0.90 (0.69-1.16) | |||||||
Swede et al. (96) USA | 1982-1998 | Hospital-based case-control study | Cases: 1,478 primary, incident cases confirmed via pathology report | Self-administered questionnaire | Regular aspirin use vs none: 0.85 (0.74-0.97) | Age at menarche, age at first birth, BMI, history of first-degree relative with breast cancer, and history of benign breast disease | |
Controls: 3,383 cancer-free controls frequency matched to cases on 5-y age intervals | ≥7 aspirin tablets/week vs none: 0.74 (0.59-0.92) | ||||||
≥10 y aspirin use vs none: 0.91 (0.78-1.06) | |||||||
Daily regular use of aspirin for ≥10 y vs none: 0.72 (0.53-0.97) | |||||||
Marshall et al. (116) USA | 1995-2001 | Cohort | Cases: 2,391 primary incident cases confirmed by tumor registry | Self-administered questionnaire, cancer registry data | Daily use aspirin vs none: 0.98 (0.86-1.13); ≥5 y regular use vs none: 1.07 (0.96-1.20) | Race, BMI, first-degree family history, menopausal and hormone therapy use status, smoking, alcohol intake, physical activity, mammography history, breast biopsy history, parity before age 30, neighborhood SES | |
114,640 disease-free cohort women | Daily use ibuprofen vs none: 1.24 (1.07-1.44); ≥5 y use vs none: 1.17 (1.00-1.36) | ||||||
Daily use any NSAID vs none: 1.09 (0.97-1.21); ≥5 y regular use vs none: 1.11 (1.01-1.23) | |||||||
Daily use acetaminophen vs none: 0.99 (0.74-1.31); ≥5 y vs none: 1.10 (0.95-1.27) | |||||||
ER/PgR negative and aspirin use ≥ 5 y daily vs none: 1.81 (1.12-2.92); ER/PgR positive and ibuprofen use ≥5 y daily vs none: 1.50 (1.11-2.03) | |||||||
Rahme et al. (129) Canada | 1998-2002 | Registry-based study | Cases: 1,090 incident cases identified from mammography screening group | Population prescription/medical record database | COX-2 inhibitors ≥90 d vs none: 0.81 (0.68-0.97) | Age, mammography in year 2 or 3 before index date, breast procedure in the prior 3 y, benign neoplasm of the breast in prior 3 y, other breast disease in the prior 3 y, HRT in prior year, visit to gynecologist in prior year | |
Controls: 44,990 disease-free women from mammography screening group (418,458 women in total cohort) | NSAID ≥90 d vs none: 0.65 (0.43-0.99) | ||||||
Aspirin > 100 mg/d for ≥90 d vs none: 0.75 (0.64-0.89) | |||||||
Acetaminophen ≥90 d vs none: 0.91 (0.71-1.16) | |||||||
Moorman et al. (108) USA | 1993-2001 | Population-based case-control study | Cases: 763 cases of invasive or in situ breast cancer among African American women | In-person interviews, genotyping by Taqman assay | COX-2 gene wild-type homozygous or heterozygous and regular NSAID use vs none: 0.3 (0.1-0.9) | Age, offset term for oversampling younger and African American women | |
Controls: 678 disease-free African American population controls matched by age | COX-2 gene variant homozygous and regular NSAID use vs none: 0.3 (0.2-0.6) | ||||||
Cook et al. (118) USA | 1992-2004 | Randomized controlled trial | 39,876 women randomized into low-dose aspirin (19,934) and placebo (19,942) arms followed for self-reported cancer endpoints verified by medical record review | Self-administered questionnaire | Aspirin use vs placebo: 0.98 (0.87-1.09) | None | |
Marginal interaction between aspirin use and smoking status, P < 0.09 | |||||||
Never smokers: 1.11 (0.94-1.30), former smokers: 0.84 (0.70-1.01), and current smokers: 0.93 (0.69-1.25) | |||||||
Harris et al. (94) USA | 2003-2004 | Hospital-based case-control study | Cases: 323 cases of histologically confirmed invasive breast cancer | In-person interview | Daily COX-2 inhibitor use ≥2 y vs none/infrequent use: 0.29 (0.14-0.59) | Age, BMI, parity, menopausal status, family history, smoking and alcohol intake | None/infrequent use defined as use of no more than one pill per week for <1 y |
Controls: 649 age, race, and residence matched controls from hospital mammography service | Aspirin use 2 times per week for ≥2 y vs none/infrequent use: 0.49 (0.26-0.94) | ||||||
Ibuprofen/naproxen use 2 times per week for ≥2 y vs none/infrequent use: 0.37 (0.18-0.72) | |||||||
Gallicchio et al. (106) USA | 1989-2003 | Cohort | Cases: 91 cases of invasive or in situ breast cancer identified via county and state cancer registries | Self-report questionnaire, medical record data, and biological sample for COX genotyping via Taqman assay | Aspirin use in 1989 vs none: 0.46 (0.22-0.98) | Age, type of NSAID | Adjustment for following variables did not appreciably effect risk estimates: education, age at menarche, menopausal status in 1989, alcohol consumption in 1989, family history of breast cancer, BMI in 1989, and parity |
1,467 women with benign breast disease identified from larger CLUE II cohort of 14,625 women | Aspirin use in 1996 vs none: 0.47 (0.18-1.21) | ||||||
Any NSAID use in 1989 vs none: 0.60 (0.35-1.03) | |||||||
Any NSAID use in 1996 vs none: 0.64 (0.32-1.27) | |||||||
No association between COX genotype and breast cancer risk | |||||||
Suggestion of significantly increased risk among those with COX-2 rs2143416 variant CC genotype and nonuse of NSAIDs | |||||||
Shen et al. (109) USA | 1996-1997 | Population-based case-control study | Cases: 1,067 in situ or invasive breast cancer cases included in the Long Island Breast Cancer Study Project | In-person interview | No major effects of the three COX-2 variant alleles on breast cancer risk were found | Age at reference (defined as age at diagnosis for cases and age at identification for controls) | Variables found not to confound associations of interest: age at menarche, parity, lactation, months of lactation, age at first birth, number of miscarriages, history of fertility problems, alcohol, race, education, religion, marital status |
Controls: 1,110 frequency matched on age identified through random-digit dialing | Among women with hormone receptor-positive breast cancer, reduced risk for any NSAID use was only evident among those who had at least one variant C allele of COX-2 0.8473 | ||||||
NSAID use vs none: 0.7 (0.5-1.0) | |||||||
P for the interaction = 0.02 | |||||||
Gill et al. (153) USA | 1993-2002 | Cohort | 1,830 breast cancer cases in the Multiethnic cohort | Self-administered questionnaire | No association between breast cancer risk and duration of aspirin use for current or past users vs nonusers was found | Age, ethnicity, BMI, family history of breast cancer, education, mammography screening, alcohol intake, age at menarche, age at fist live birth, number of children, menopausal status, and HRT | |
98,920 women in cohort | Duration of current other NSAID use protective vs nonusers (≥6 y): 0.70 (0.51-0.95) | ||||||
When stratified by ethnicity and hormone receptor status, the protective effect limited to Caucasians or African Americans or to women with at least one positive hormone receptor | |||||||
Jacobs et al. (119) USA | 1992-2003 | Cohort | 571 breast cancer cases in Cancer Prevention Study II Nutrition cohort | Self-administered questionnaire | Less than daily, low-dose, or past use vs no reported use: 1.10 (1.00-1.21) | Age, race, education, smoking, BMI, physical activity level, use of HRT, history of mammography, history of colorectal endoscopy, use of use of nonaspirin NSAIDs, history of heart attack, diabetes, hypertension | Adjustment for following variables did not appreciably effect risk estimates: nutritional factors |
76,303 total women in cohort | Not statistically significant lower risk for current daily use (≥325 mg) ≥5 y: 0.83 (0.63-1.10) | ||||||
Ready et al. (114) USA | 2000-2002 | Cohort | 482 breast cancer cases in VITAL cohort study | Self-administered questionnaire | Low-dose aspirin overall use vs none: 0.99 (0.80-1.23) | Age, race, BMI, family history of breast cancer, history of biopsy, mammogram in 2 y before baseline, age at menarche, age at first birth, age at menopause, history of surgical menopause, years of combined estrogen and progesterone hormone therapy, multivitamin use and alcohol use, adjustment for use of other categories of NSAIDs | |
35,323 total postmenopausal women in cohort | Low-dose aspirin at ≥4 d/wk over 10 y vs none: 0.65 (0.43-0.91) | ||||||
All NSAIDs (except for low-dose aspirin) overall use vs none: 0.98 (0.67-1.44) | |||||||
All NSAIDs (except for low-dose aspirin) ≤1-3 d/wk over 10 y vs none: 0.78 (0.61-0.98) | |||||||
All NSAIDs (except for low-dose aspirin) ≥4 d/wk over 10 y vs none: 1.26 (0.96-1.65) | |||||||
Regular/extra-strength aspirin at ≥4 d/wk over 10 y vs none: 1.43 (1.02-2.00) | |||||||
Gallicchio et al. (107) USA | 1989-2006 | Cohort | 430 cases of primary invasive breast cancer identified from cancer registries | In-person interview | Nonaspirin NSAID use in 1996 vs nonusers: 0.53 (0.31-0.93) | Age at baseline | Adjustment for following variables did not appreciably affect risk estimates: education, history of fibrocystic disease, family history of breast cancer, age at first menarche, hormone use, oral contraceptive use, menopausal status, parity, BMI |
18,723 total women in cohort | NSAID use at baseline and in 1996 vs no NSAID use at baseline and in 1996: 0.50 (0.28-0.91) | ||||||
Bardia et al. (112) USA | 1992-2003 | Cohort | 3,487 incident cancer cases and 3,581 deaths were observed in the cohort of 22,507 postmenopausal women | Self-administered questionnaire | Aspirin use vs nonuse was inversely associates with total cancer incidence: 0.84 (0.77-0.90) or cancer mortality: 0.87 (0.76-0.99) | Age, education status, physical activity, use of HRT, marital status, BMI, diabetes status, fruit and vegetable intake, waist-to-hip ratio, history of hypertension, alcohol use, vitamin supplement use, total caloric intake, red meat consumption, whole-wheat consumption, vitamin E intake, cholesterol intake, history of osteoarthritis, and history of rheumatoid arthritis | No information according to sites |
The inverse relationship was stronger among former and never smokers vs current smokers | |||||||
Nonaspirin NSAID use was not associated with cancer incidence or mortality | |||||||
Davis and Mirick (103) USA | 1992-1995 | Population-based case-control study | Cases: 600 newly diagnosed breast cancer | Telephone interview | No association between risk of breast cancer and any measure of NSAID use | Parity, age at first pregnancy, mother/sister breast cancer, early double oophorectomy, oral contraceptive use, ever upper gastrointestinal series, and ever smoker (all subjects); mother/sister breast cancer ages <45 y and alcohol intake (if premenopausal) or HRT (if postmenopausal) | |
Controls: 647 from the Seattle metropolitan area, identified by random-digit dialing and frequency matched by 5-y age groups | Ever regular NSAID use vs never: 1.1 (0.8-1.4) | ||||||
<5 y vs never: 1.1 (0.7-1.8) | |||||||
5-10 y vs never: 1.0 (0.7-1.5) | |||||||
≥2 y before diagnosis vs never: 2.0 (0.9-4.3) | |||||||
<2 y diagnosis vs never: 1.0 (0.7-1.3) | |||||||
Among cases with localized disease, ≥2 y before diagnosis vs never: 2.2 (1.0-4.9) | |||||||
Slattery et al. (101) | 1999-2004 | Population-based case-control study | Cases: 798 Hispanic/Native American and 1,527 non-Hispanic White women diagnosed with first primary breast cancer | In-person interview | Aspirin use vs nonuse among postmenopausal women with no recent hormone exposure: 0.56 (0.33-0.96) | Age, study center, referent year BMI, lifetime physical activity score, parity, and percentage Native American ancestry | |
Controls: 935 Hispanic/Native American and 1,671 non-Hispanic White women from the target populations matched on ethnicity and 5-year age distribution of cases | Aspirin use among postmenopausal women with recent hormone exposure or premenopausal/perimenopausal women was not associated with breast cancer risk | ||||||
Interleukin-6 genotype modified the association between aspirin and breast cancer among postmenopausal women with no recent hormone exposure (P for interaction = 0.04 for Hispanic/Native American and 0.06 for non-Hispanic White) | |||||||
Friis et al. (82) Denmark | 1993-2003 | Cohort | 847 cases identified via the Danish Cancer Registry | Self-administered questionnaire at baseline (1993-1997) and data updated using a nationwide prescription database through 2-3 | Any NSAID use at baseline vs nonuse: 1.27 (1.10-1.45) | Age, school education, parity number of births, use of HRT, and history of benign breast tumor surgery | |
29,875 total cohort member | Similar results were observed in a combined analysis of baseline and prescription data | ||||||
Aspirin only use vs nonuse: 1.31 (1.12-1.53) | |||||||
No differences in risk estimates with frequency, recency, or duration of NSAID use or by hormone receptor status of breast tumors | |||||||
Mangiapane et al. (121) N/A | N/A | Meta-analysis | N/A | Literature database search from 2001 to 2005 for cohort or case-control studies | Aspirin use vs none: 0.74 (0.69-0.79) in all studies, 0.82 (0.73-0.92) in 4 cohort studies, and 0.70 (0.56-0.87) in 6 case-control studies | N/A | |
10 studies were analyzed | |||||||
Zhang et al. (83) USA | 1992-2004 | Randomized controlled trial | 39,876 women randomized into low-dose aspirin (19,934) and placebo (19,942) arms followed for self-reported cancer endpoints verified by medical record review | Self-administered questionnaire | Low-dose aspirin has no preventive effect of breast cancer in the subgroup analysis by tumor characteristics at diagnosis | None |
Study and country . | Years . | Design . | Cases/controls . | Exposure measurement . | Major findings . | Confounders . | Comments . |
---|---|---|---|---|---|---|---|
Schreinemachers and Everson (110) USA | 1971-1975 | Cohort | 1,257 cases identified via the National Health and Examination Survey I | In-person and telephone interviews, hospital and nursing home records | Incident risk ratio for all sites combined for aspirin users vs nonaspirin users (30 d before interview) | Gender, age | Adjustment for following variables did not appreciably effect incident risk ratio: race, education, smoking, alcohol |
11,411 cancer-free cohort members | All sites combined: 0.83 (0.74-0.93), lung cancer: 0.68 (0.49-0.94), breast cancer in women: 0.70 (0.50-0.96), and colorectal cancer in younger men: 0.35 (0.17-0.73) | ||||||
Harris et al. (95) USA | 1988-1992 | Hospital-based case-control study | Cases: 744 patients with newly diagnosed breast cancer identified by collaborating hospitals in northeastern United States | In-person interview | 1-4 y NSAID use vs none: 1.09 (0.8-1.5), ≥5 y NSAID use vs none: 0.63 (0.5-0.9) | Age, menopausal status, parity, family history of breast cancer, BMI | |
Controls: 767 patients without cancer diagnoses frequency matched to cases | |||||||
Egan et al. (117) USA | 1980-1992 | Cohort | 2,414 cases of invasive breast cancer (2,303 confirmed with medical records and 111 cases identified by questionnaire response) | Self-administered questionnaire, medical record review | Regular aspirin use from 1980 to 1988 vs no regular use: 1.01 (0.80-1.27) | Age at menarche, age at menopause, BMI, alcohol, family history of breast cancer, history of benign breast disease, multivitamin use | Authors concluded that regular aspirin use does not reduce breast cancer risk |
Heavy use from 1980 to 1988 vs no regular use: 1.09 (0.75-1.60) | |||||||
≥20 y regular use vs no regular use: 1.00 (0.71-1.41) | |||||||
Harris et al. (99) USA | Not noted | Population-based case-control study | Cases: 511 newly diagnosed breast cancer confirmed by pathology report | In-person interview | Regular NSAID use vs no regular use: 0.66 (0.52-0.83) | Age, parity, menopausal status, family history | |
Controls: 1,534 cancer-free women from central OH frequency matched by race and age | Regular aspirin use vs no regular use: 0.69 (0.46-0.99) | ||||||
Regular ibuprofen use vs no regular use: 0.57 (0.36-0.91) | |||||||
≥7 per week, ≥5 y NSAID use vs no regular use: 0.60 (0.40-0.91) | |||||||
Harris et al. (113) USA | 1991-1993 | Cohort | 393 breast cancers have been detected | Self-administered questionnaire | 1-3 NSAID pills per week vs <1: 0.64 (0.50-0.82), ≥4 vs <1: 0.57 (0.44-0.74) | None | Adjustment for the following variables did not appreciably effect risk estimates: age, education, parity, menopausal status, and family history of breast cancer |
32,505 women enrolled in the mammography screening program of The Ohio State University Comprehensive Cancer Center (4.7 y average follow-up) | |||||||
Coogan et al. (65) USA | 1976-1996 | Hospital-based case-control study | Cases: 6,558 women with a first occurrence of primary breast cancer diagnosed within the previous year, confirmed by path report, and no concurrent or previous cancer | In-person interview | For cancer controls: regular use within 1 y of admission only vs never use: 0.6 (0.4-1.0); regular use begun ≥1 y before admission vs never use: 0.8 (0.7-1.0) | Age, study center, interview year, years of education, history of benign breast disease, number of doctor visits 2 y before admission, duration of oral contraceptive use, duration of use of female hormone supplements | Adjustment for the following variables did not appreciably effect risk estimates: age at menarche, age at menopause, age at first birth, parity, race, alcohol consumption, religion, breast cancer in mother or sister, practice of breast self-examination, BMI |
Controls: 3,296 patients with other cancers not associated with NSAID use, 2,925 noncancer patients | For noncancer controls: regular use within 1 y of admission only vs never use: 0.5 (0.3-0.8), regular use begun ≥1 y before admission vs never use: 0.7 (0.6-0.9) | ||||||
Sharpe et al. (93) Canada | 1981-1995 | Registry-based study | Cases: 5,882 women diagnosed with histologically proven invasive breast cancer | Saskatchewan Prescription Drug Plan database | NSAID exposure 2-5 y before diagnosis: average daily dose >0.3 vs ADD = 0: 0.76 (0.63-0.92) | Sampling fractions, age, exposure during other periods, total duration of lactation, BMI after menopause | For periods <2 and >5 y, there was no significant reduction in risk |
Controls: 23,517 controls frequency matched on age and sampling time | Just cases, exposure 2-5 y before diagnosis: 0 < ADD ≤ 0.1 vs ADD = 0: 0.52 (0.37-0.73), 0.1 < ADD ≤ 0.3 vs ADD = 0: 0.53 (0.30-0.92), and ADD > 0.3 vs ADD = 0: 0.49 (0.24-0.99) | ||||||
Khuder and Mutgi (120) N/A | N/A | Meta-analysis | N/A | Literature database search through 2000 | NSAID use vs none | N/A | The numbers given in Table 3 are different than the numbers presented in the body of the article |
14 articles were analyzed | 0.82 (0.75-0.89) in all studies, 0.78 (0.62-0.99) in 6 cohort studies, and 0.87 (0.84-0.91) in 8 case-control studies | ||||||
Cotterchio et al. (98) Canada | 1996-1998 | Population-based case-control study | Cases: 3,133 random women diagnosed with a first primary cancer of the breast, 25-74 y identified via Ontario Cancer Registry | Self-report questionnaire | Any regular NSAID use vs never: 0.76 (0.66-0.88) | Age, history of arthritis, benign breast disease | Confounders evaluated include HRT, oral contraceptive use, alcohol, smoking, weight, BMI, physical activity, history of arthritis, reproductive history, education, marital status, previous breast cysts, family history of breast cancer, other medication use, dietary fat intake |
Controls: 3,062 age-matched random sample of the female population of Ontario | ≥9 y NSAID use vs never: 0.68 (0.54-0.86) | ||||||
≤1 y since last NSAID use vs never: 0.64 (0.54-0.77) | |||||||
Age at first use ≥50 vs never: 0.76 (0.61-0.93) | |||||||
Meier et al. (128) UK | Registry-based study | Cases: 3,706 women with incident breast cancer | Medical history computer record | 20-29 acetaminophen prescriptions vs none: 0.7 (0.6-0.9) | Smoking status, BMI | Adjustment for following variables did not appreciably effect risk estimates: prior hysterectomy, prior oophorectomy, prior history of benign breast lumps, longer-term exposure to postmenopausal estrogens | |
Controls: 14,155 age, years of medical history in the computer record, general practice attended, and calendar time matched controls | ≥30 acetaminophen prescriptions vs none: 0.8 (0.7-1.0) | Study contains no information about over-the-counter NSAID | |||||
No statistically significant difference found between number of NSAID prescriptions | |||||||
Johnson et al. (111) USA | 1986-1997 | Cohort | 938 cases identified from the Iowa Women's Health Study | Self-administered questionnaire | Aspirin or NSAID use vs none: 0.80 (0.67-0.95) | Age, BMI, estrogen use, family history of breast cancer, benign breast disease, multivitamin use, category of NSAID use, mammography, waist-to-hip ratio | |
27,616 total cohort members | Aspirin use vs none: 0.82 (0.71-0.95) | ||||||
NSAID use vs none: 0.98 (0.85-1.14) | |||||||
≥6 aspirin use per week vs none: 0.71 (0.58-0.87) | |||||||
In situ disease: ≥6 aspirin use vs none: 0.52 (0.30, 0.90) | |||||||
Regional or distant disease: ≥6 aspirin use vs none: 0.50 (0.29-0.88) | |||||||
Harris et al. (105) USA | Not noted | Cohort study | 1,392 self-reported incident cases confirmed by medical record review | Self-administered questionnaire | Any NSAID use for ≥10 vs <1 y use: 0.72 (0.56-0.91) | Age, ethnicity, education, BMI, HRT use, family history of breast cancer, parity at age <30 y, and episodes of weekly exercise | Additional analyses stratified by BMI, HRT use, family history of breast cancer, parous at age <30 y, and episodes per week of moderate/strenuous exercise did not vary appreciably from full sample |
80,741 women in total cohort (43-mo average follow-up) | Aspirin use for ≥10 vs <1 y use: 0.79 (0.60-1.03) | ||||||
Ibuprofen use for ≥10 vs <1 y use: 0.51 (0.28-0.96) | |||||||
Moorman et al. (100) USA | 1996-2000 | Population-based case-control study | Cases: 930 cases of invasive breast cancer identified via North Carolina Central Cancer Registry | In-person interview | Any NSAID use vs none: 0.4 (0.3-0.6) | Age, race, age at menarche, age at first full-term pregnancy, breastfeeding history, menopausal status, family history, oral contraceptive use, HRT use, education, BMI, waist-to-hip ratio, smoking status, and offset term | |
Controls: 754 controls selected from DMV and Health Care Financing Administration, frequency matched to cases on age and ethnicity | Occasional NSAID use vs none: 0.5 (0.3-0.7) | ||||||
Regular NSAID use ≥3 y vs none: 0.3 (0.2-0.5) | |||||||
Gonzalez-Perez et al. (122) N/A | N/A | Meta-analysis | N/A | Literature database search from 1966 to 2002 for cohort or case-control studies | NSAID use vs none: 0.77 (0.66-0.88) | N/A | |
15 of 47 studies reporting outcomes for breast cancer were analyzed | Aspirin use vs none: 0.77 (0.69-0.86) | ||||||
Garcia Rodriguez and Gonzalez-Perez (104) UK | 1995-2001 | Registry-based study | Cases: 3,708 cases of invasive breast cancer identified from General Practice Research | General practice database/electronic medical record | Aspirin use ≥4 y vs none: 0.86 (0.61-1.19) | Age, calendar year, BMI, alcohol intake, smoking status, HRT use, prior benign breast disease, and remaining NSAID | |
Database | Nonaspirin NSAID (ibuprofen) use ≥4 y vs none: 0.94 (0.74-1.21) | ||||||
Controls: 20,000 cancer-free controls from cohort matched to cases on age and calendar year (study cohort = 734,899 women) | Acetaminophen use ≥4 y vs none: 0.77 (0.64-0.94) | ||||||
Terry et al. (102) USA | 1996-1997 | Population-based case-control study | Cases: 1,508 invasive or in situ breast cancer cases confirmed by medical record review | In-person interviews, medical records | Aspirin use ≥7 times/wk for ≥5 y vs none: 0.77 (0.57-1.04) | Age at diagnosis, migraine headache, BMI | |
Controls: 1,556 controls selected though random-digit dialing methods and Health Care Financing Administration lists, frequency matched to cases in 5-y age intervals | Ibuprofen use ≥3 times/wk for ≥5 y vs none: 1.09 (0.70-1.70) | ||||||
Ever used aspirin and hormone receptor positive vs none: 0.74 (0.60-0.93) | |||||||
Ever used aspirin and hormone receptor negative vs none: 0.97 (0.67-1.40) | |||||||
Harris et al. (160) N/A | N/A | Meta-analysis | N/A | Literature database search through 1970-2003 | All studies NSAID vs no use: RR, 0.61 (0.50-0.75) | N/A | |
17 studies | |||||||
Jacobs et al. (115) USA | 1992-2001 | Cohort study | 3008 incident cases identified via self-report and confirmed via medical record or state cancer registries | Self-administered questionnaire | ≥60 tablets of any NSAID per month vs none: 1.07 (0.96-1.21) | Age, race, education, family history of breast cancer, personal history of breast cysts, history of mammography, age at menarche, duration of oral contraceptive use, parity, age at menopause, HRT use, weight change, BMI, alcohol consumption | |
97,786 women in total cohort | ≥60 aspiring tablets per month vs none: 1.01 (0.84-1.20) | ||||||
≥60 ibuprofen tablets per month vs none: 1.06 (0.89-1.26) | |||||||
≥5 y current regular NSAID use vs none: 1.05 ((0.88-1.26) | |||||||
≥5 y current regular aspirin use vs none: 0.88 (0.69-1.12) | |||||||
≥5 y current regular ibuprofen use vs none: 1.29 (0.92-1.82) | |||||||
Zhang et al. (97) USA | 1976-2002 | Hospital-based case-control study | Cases: 7,006 primary breast cancer cases confirmed from discharge summaries and pathology reports | In-person interview | ≥20 y regular aspirin use vs none: 0.59 (0.25-1.36) | Age, year of interview, study center, race, year of education, benign breast disease, number of physician visits 2 y before hospitalization, duration of HRT use, duration of oral contraceptive use, age at menarche, age at menopause, age at first birth, parity, alcohol consumption, family history of breast cancer, practice of breast self-exam, and BMI | |
Controls: 3,622 controls admitted for nonmalignant conditions | ≥5 y regular ibuprofen use v none: 0.78 (0.29-2.08) | ||||||
Regular use of aspirin and hormone receptor positive vs none: 0.74 (0.44-1.26) | |||||||
Regular use of aspirin and hormone receptor negative vs none: 0.94 (0.45-1.96) | |||||||
Regular use any NSAID and premenopausal vs none: 0.62 (0.41-0.94) | |||||||
Regular use any NSAID and postmenopausal vs none: 0.90 (0.69-1.16) | |||||||
Swede et al. (96) USA | 1982-1998 | Hospital-based case-control study | Cases: 1,478 primary, incident cases confirmed via pathology report | Self-administered questionnaire | Regular aspirin use vs none: 0.85 (0.74-0.97) | Age at menarche, age at first birth, BMI, history of first-degree relative with breast cancer, and history of benign breast disease | |
Controls: 3,383 cancer-free controls frequency matched to cases on 5-y age intervals | ≥7 aspirin tablets/week vs none: 0.74 (0.59-0.92) | ||||||
≥10 y aspirin use vs none: 0.91 (0.78-1.06) | |||||||
Daily regular use of aspirin for ≥10 y vs none: 0.72 (0.53-0.97) | |||||||
Marshall et al. (116) USA | 1995-2001 | Cohort | Cases: 2,391 primary incident cases confirmed by tumor registry | Self-administered questionnaire, cancer registry data | Daily use aspirin vs none: 0.98 (0.86-1.13); ≥5 y regular use vs none: 1.07 (0.96-1.20) | Race, BMI, first-degree family history, menopausal and hormone therapy use status, smoking, alcohol intake, physical activity, mammography history, breast biopsy history, parity before age 30, neighborhood SES | |
114,640 disease-free cohort women | Daily use ibuprofen vs none: 1.24 (1.07-1.44); ≥5 y use vs none: 1.17 (1.00-1.36) | ||||||
Daily use any NSAID vs none: 1.09 (0.97-1.21); ≥5 y regular use vs none: 1.11 (1.01-1.23) | |||||||
Daily use acetaminophen vs none: 0.99 (0.74-1.31); ≥5 y vs none: 1.10 (0.95-1.27) | |||||||
ER/PgR negative and aspirin use ≥ 5 y daily vs none: 1.81 (1.12-2.92); ER/PgR positive and ibuprofen use ≥5 y daily vs none: 1.50 (1.11-2.03) | |||||||
Rahme et al. (129) Canada | 1998-2002 | Registry-based study | Cases: 1,090 incident cases identified from mammography screening group | Population prescription/medical record database | COX-2 inhibitors ≥90 d vs none: 0.81 (0.68-0.97) | Age, mammography in year 2 or 3 before index date, breast procedure in the prior 3 y, benign neoplasm of the breast in prior 3 y, other breast disease in the prior 3 y, HRT in prior year, visit to gynecologist in prior year | |
Controls: 44,990 disease-free women from mammography screening group (418,458 women in total cohort) | NSAID ≥90 d vs none: 0.65 (0.43-0.99) | ||||||
Aspirin > 100 mg/d for ≥90 d vs none: 0.75 (0.64-0.89) | |||||||
Acetaminophen ≥90 d vs none: 0.91 (0.71-1.16) | |||||||
Moorman et al. (108) USA | 1993-2001 | Population-based case-control study | Cases: 763 cases of invasive or in situ breast cancer among African American women | In-person interviews, genotyping by Taqman assay | COX-2 gene wild-type homozygous or heterozygous and regular NSAID use vs none: 0.3 (0.1-0.9) | Age, offset term for oversampling younger and African American women | |
Controls: 678 disease-free African American population controls matched by age | COX-2 gene variant homozygous and regular NSAID use vs none: 0.3 (0.2-0.6) | ||||||
Cook et al. (118) USA | 1992-2004 | Randomized controlled trial | 39,876 women randomized into low-dose aspirin (19,934) and placebo (19,942) arms followed for self-reported cancer endpoints verified by medical record review | Self-administered questionnaire | Aspirin use vs placebo: 0.98 (0.87-1.09) | None | |
Marginal interaction between aspirin use and smoking status, P < 0.09 | |||||||
Never smokers: 1.11 (0.94-1.30), former smokers: 0.84 (0.70-1.01), and current smokers: 0.93 (0.69-1.25) | |||||||
Harris et al. (94) USA | 2003-2004 | Hospital-based case-control study | Cases: 323 cases of histologically confirmed invasive breast cancer | In-person interview | Daily COX-2 inhibitor use ≥2 y vs none/infrequent use: 0.29 (0.14-0.59) | Age, BMI, parity, menopausal status, family history, smoking and alcohol intake | None/infrequent use defined as use of no more than one pill per week for <1 y |
Controls: 649 age, race, and residence matched controls from hospital mammography service | Aspirin use 2 times per week for ≥2 y vs none/infrequent use: 0.49 (0.26-0.94) | ||||||
Ibuprofen/naproxen use 2 times per week for ≥2 y vs none/infrequent use: 0.37 (0.18-0.72) | |||||||
Gallicchio et al. (106) USA | 1989-2003 | Cohort | Cases: 91 cases of invasive or in situ breast cancer identified via county and state cancer registries | Self-report questionnaire, medical record data, and biological sample for COX genotyping via Taqman assay | Aspirin use in 1989 vs none: 0.46 (0.22-0.98) | Age, type of NSAID | Adjustment for following variables did not appreciably effect risk estimates: education, age at menarche, menopausal status in 1989, alcohol consumption in 1989, family history of breast cancer, BMI in 1989, and parity |
1,467 women with benign breast disease identified from larger CLUE II cohort of 14,625 women | Aspirin use in 1996 vs none: 0.47 (0.18-1.21) | ||||||
Any NSAID use in 1989 vs none: 0.60 (0.35-1.03) | |||||||
Any NSAID use in 1996 vs none: 0.64 (0.32-1.27) | |||||||
No association between COX genotype and breast cancer risk | |||||||
Suggestion of significantly increased risk among those with COX-2 rs2143416 variant CC genotype and nonuse of NSAIDs | |||||||
Shen et al. (109) USA | 1996-1997 | Population-based case-control study | Cases: 1,067 in situ or invasive breast cancer cases included in the Long Island Breast Cancer Study Project | In-person interview | No major effects of the three COX-2 variant alleles on breast cancer risk were found | Age at reference (defined as age at diagnosis for cases and age at identification for controls) | Variables found not to confound associations of interest: age at menarche, parity, lactation, months of lactation, age at first birth, number of miscarriages, history of fertility problems, alcohol, race, education, religion, marital status |
Controls: 1,110 frequency matched on age identified through random-digit dialing | Among women with hormone receptor-positive breast cancer, reduced risk for any NSAID use was only evident among those who had at least one variant C allele of COX-2 0.8473 | ||||||
NSAID use vs none: 0.7 (0.5-1.0) | |||||||
P for the interaction = 0.02 | |||||||
Gill et al. (153) USA | 1993-2002 | Cohort | 1,830 breast cancer cases in the Multiethnic cohort | Self-administered questionnaire | No association between breast cancer risk and duration of aspirin use for current or past users vs nonusers was found | Age, ethnicity, BMI, family history of breast cancer, education, mammography screening, alcohol intake, age at menarche, age at fist live birth, number of children, menopausal status, and HRT | |
98,920 women in cohort | Duration of current other NSAID use protective vs nonusers (≥6 y): 0.70 (0.51-0.95) | ||||||
When stratified by ethnicity and hormone receptor status, the protective effect limited to Caucasians or African Americans or to women with at least one positive hormone receptor | |||||||
Jacobs et al. (119) USA | 1992-2003 | Cohort | 571 breast cancer cases in Cancer Prevention Study II Nutrition cohort | Self-administered questionnaire | Less than daily, low-dose, or past use vs no reported use: 1.10 (1.00-1.21) | Age, race, education, smoking, BMI, physical activity level, use of HRT, history of mammography, history of colorectal endoscopy, use of use of nonaspirin NSAIDs, history of heart attack, diabetes, hypertension | Adjustment for following variables did not appreciably effect risk estimates: nutritional factors |
76,303 total women in cohort | Not statistically significant lower risk for current daily use (≥325 mg) ≥5 y: 0.83 (0.63-1.10) | ||||||
Ready et al. (114) USA | 2000-2002 | Cohort | 482 breast cancer cases in VITAL cohort study | Self-administered questionnaire | Low-dose aspirin overall use vs none: 0.99 (0.80-1.23) | Age, race, BMI, family history of breast cancer, history of biopsy, mammogram in 2 y before baseline, age at menarche, age at first birth, age at menopause, history of surgical menopause, years of combined estrogen and progesterone hormone therapy, multivitamin use and alcohol use, adjustment for use of other categories of NSAIDs | |
35,323 total postmenopausal women in cohort | Low-dose aspirin at ≥4 d/wk over 10 y vs none: 0.65 (0.43-0.91) | ||||||
All NSAIDs (except for low-dose aspirin) overall use vs none: 0.98 (0.67-1.44) | |||||||
All NSAIDs (except for low-dose aspirin) ≤1-3 d/wk over 10 y vs none: 0.78 (0.61-0.98) | |||||||
All NSAIDs (except for low-dose aspirin) ≥4 d/wk over 10 y vs none: 1.26 (0.96-1.65) | |||||||
Regular/extra-strength aspirin at ≥4 d/wk over 10 y vs none: 1.43 (1.02-2.00) | |||||||
Gallicchio et al. (107) USA | 1989-2006 | Cohort | 430 cases of primary invasive breast cancer identified from cancer registries | In-person interview | Nonaspirin NSAID use in 1996 vs nonusers: 0.53 (0.31-0.93) | Age at baseline | Adjustment for following variables did not appreciably affect risk estimates: education, history of fibrocystic disease, family history of breast cancer, age at first menarche, hormone use, oral contraceptive use, menopausal status, parity, BMI |
18,723 total women in cohort | NSAID use at baseline and in 1996 vs no NSAID use at baseline and in 1996: 0.50 (0.28-0.91) | ||||||
Bardia et al. (112) USA | 1992-2003 | Cohort | 3,487 incident cancer cases and 3,581 deaths were observed in the cohort of 22,507 postmenopausal women | Self-administered questionnaire | Aspirin use vs nonuse was inversely associates with total cancer incidence: 0.84 (0.77-0.90) or cancer mortality: 0.87 (0.76-0.99) | Age, education status, physical activity, use of HRT, marital status, BMI, diabetes status, fruit and vegetable intake, waist-to-hip ratio, history of hypertension, alcohol use, vitamin supplement use, total caloric intake, red meat consumption, whole-wheat consumption, vitamin E intake, cholesterol intake, history of osteoarthritis, and history of rheumatoid arthritis | No information according to sites |
The inverse relationship was stronger among former and never smokers vs current smokers | |||||||
Nonaspirin NSAID use was not associated with cancer incidence or mortality | |||||||
Davis and Mirick (103) USA | 1992-1995 | Population-based case-control study | Cases: 600 newly diagnosed breast cancer | Telephone interview | No association between risk of breast cancer and any measure of NSAID use | Parity, age at first pregnancy, mother/sister breast cancer, early double oophorectomy, oral contraceptive use, ever upper gastrointestinal series, and ever smoker (all subjects); mother/sister breast cancer ages <45 y and alcohol intake (if premenopausal) or HRT (if postmenopausal) | |
Controls: 647 from the Seattle metropolitan area, identified by random-digit dialing and frequency matched by 5-y age groups | Ever regular NSAID use vs never: 1.1 (0.8-1.4) | ||||||
<5 y vs never: 1.1 (0.7-1.8) | |||||||
5-10 y vs never: 1.0 (0.7-1.5) | |||||||
≥2 y before diagnosis vs never: 2.0 (0.9-4.3) | |||||||
<2 y diagnosis vs never: 1.0 (0.7-1.3) | |||||||
Among cases with localized disease, ≥2 y before diagnosis vs never: 2.2 (1.0-4.9) | |||||||
Slattery et al. (101) | 1999-2004 | Population-based case-control study | Cases: 798 Hispanic/Native American and 1,527 non-Hispanic White women diagnosed with first primary breast cancer | In-person interview | Aspirin use vs nonuse among postmenopausal women with no recent hormone exposure: 0.56 (0.33-0.96) | Age, study center, referent year BMI, lifetime physical activity score, parity, and percentage Native American ancestry | |
Controls: 935 Hispanic/Native American and 1,671 non-Hispanic White women from the target populations matched on ethnicity and 5-year age distribution of cases | Aspirin use among postmenopausal women with recent hormone exposure or premenopausal/perimenopausal women was not associated with breast cancer risk | ||||||
Interleukin-6 genotype modified the association between aspirin and breast cancer among postmenopausal women with no recent hormone exposure (P for interaction = 0.04 for Hispanic/Native American and 0.06 for non-Hispanic White) | |||||||
Friis et al. (82) Denmark | 1993-2003 | Cohort | 847 cases identified via the Danish Cancer Registry | Self-administered questionnaire at baseline (1993-1997) and data updated using a nationwide prescription database through 2-3 | Any NSAID use at baseline vs nonuse: 1.27 (1.10-1.45) | Age, school education, parity number of births, use of HRT, and history of benign breast tumor surgery | |
29,875 total cohort member | Similar results were observed in a combined analysis of baseline and prescription data | ||||||
Aspirin only use vs nonuse: 1.31 (1.12-1.53) | |||||||
No differences in risk estimates with frequency, recency, or duration of NSAID use or by hormone receptor status of breast tumors | |||||||
Mangiapane et al. (121) N/A | N/A | Meta-analysis | N/A | Literature database search from 2001 to 2005 for cohort or case-control studies | Aspirin use vs none: 0.74 (0.69-0.79) in all studies, 0.82 (0.73-0.92) in 4 cohort studies, and 0.70 (0.56-0.87) in 6 case-control studies | N/A | |
10 studies were analyzed | |||||||
Zhang et al. (83) USA | 1992-2004 | Randomized controlled trial | 39,876 women randomized into low-dose aspirin (19,934) and placebo (19,942) arms followed for self-reported cancer endpoints verified by medical record review | Self-administered questionnaire | Low-dose aspirin has no preventive effect of breast cancer in the subgroup analysis by tumor characteristics at diagnosis | None |
In contrast, initial analyses from the Cancer Prevention Study II Nutrition cohort (115) as well as results from the California Teachers (116) and Nurses' Health Study (117) cohorts did not show associations between use of aspirin or other NSAIDs and breast cancer risk. In fact, in the California Teachers cohort, prolonged use (≥5 years) of both aspirin and ibuprofen was associated with significant risk elevations for women with hormone receptor-negative tumors (RR, 1.8; 95% CI, 1.12-2.92 and RR, 1.50; 95% CI, 1.1-2.03, respectively). The Danish Diet, Cancer and Health cohort study (82) also showed increased breast cancer incidence among both any NSAID and aspirin-only users (RR, 1.27; 95% CI, 1.10-1.45 and RR, 1.31; 95% CI, 1.12-1.53, respectively), although this cohort women had higher breast cancer incidence than women in the general Danish population and most chronic aspirin use came from low-dose aspirin. In the Multiethnic cohort [153], authors observed no association between aspirin and breast cancer but found that current other NSAID use was protective among Caucasian and African American as well as among women with at least one positive hormone receptor. In a randomized low-dose aspirin (100 mg) chemoprevention trial (118), with an average of 10 years of follow-up, women who were randomized to the aspirin intervention arm were not at lower risk of breast cancer compared with women who received the placebo (RR, 0.98; 95% CI, 0.87-1.09). In subgroup analyses, low-dose aspirin showed no effects by tumor characteristics at diagnosis (83) but suggested protective effects by smoking status (RR, 0.84; 95% CI, 0.70-1.01; ref. 118). Consistently, the Iowa Women's Health Study showed that the inverse association between total cancer incidence (and mortality) and aspirin use was stronger among former and never smokers than current smokers (112). However, results from the Women's Health Study, a randomized prevention trial, did not reveal lower risk of breast cancer in the treatment group after an average of 10 years of follow-up of almost 40,000 women (83, 118). It should be noted, though, that low-dose aspirin (100 mg every other day) was administered in this trial. Jacobs et al. (119) conducted further analyses in the Cancer Prevention Study II Nutrition cohort and focused on long-term (≥5 years) daily use of adult-strength aspirin preparations (≥325 mg). The authors speculated that the lack of a protective effect in the randomized trial might be due to the administration of low-dose aspirin tablets, which may not have been sufficient to produce a chemoprotective effect. Results indicated that daily long-term use was associated with a nonsignificant risk reduction (RR, 0.83; 95% CI, 0.63-1.10).
Finally, four meta-analyses showed significant chemopreventive effects of aspirin or NSAIDs against breast cancer. The first considered 14 studies published until 2000 (120) and showed a significant risk reduction associated with NSAID use (OR, 0.82; 95% CI, 0.75-0.89). A more recent meta-analysis restricted to 10 epidemiologic studies published from 2001 to 2005 (121) supported a protective association between aspirin intake and breast cancer (RR, 0.74; 95% CI, 0.69-0.79) with significant dose-response relationship. The protective effect was similar when cohort and case-control studies were examined separately (120, 121). Similar results were observed in two literature-based meta-analyses (122, 160).
Most observational studies and meta-analyses showed consistent and statistically significant risk reductions in human breast cancer with exposure to NSAIDs; however, interpretation of the existing body of literature on the associations between various NSAIDs and breast cancer risk is not straightforward. Although most studies on this topic have shown statistically significant risk reductions, the majority of these studies were either registry-based or employed a case-control design. The former approach is methodologically limited due to insufficient adjustment for potential confounders, whereas the latter study design is known to be prone to selection and information bias. Further, studies using only prescription records or health plan data will misclassify over-the-counter medication users as unexposed and thereby may underestimate exposure prevalence. Four large follow-up studies (82, 115, 150, 151) found no evidence of reduced risk of breast cancer among aspirin users, yet the majority of cohort studies found significant risk reductions among aspirin users (83, 139-141, 144-146, 148, 153). Importantly, however, two randomized trials, considered the gold standard in epidemiologic study designs, did not show a chemoprotective effect of aspirin use. It is possible, as suggested by Jacobs et al. (119) that higher-dose aspirin preparations may be needed to produce a chemoprotective effect. However, because they are the most common cause of serious gastrointestinal complications in the United States (161-163), chemopreventive trial of adult-dose (e.g., 325 mg) aspirin might be problematic. It is also possible that selective COX-2 inhibitors have much stronger chemopreventive properties than aspirin. Although previous trials revealed the serious side effects related to cardiovascular events with these drugs (164-166), recent reviews and meta-analyses of controlled observational studies (167) and randomized trials (123) confirmed that only rofecoxib was associated with the risk of cardiovascular events and suggests that celecoxib and other COX-2 inhibitors in commonly used doses may not increase the risk. Thus, additional randomized trials with these COX-2 inhibitors may be needed to resolve these questions. In conclusion, although the lack of a protective effect of aspirin in randomized trials is somewhat worrisome, the overwhelming majority of the existing evidence points to a chemoprotective role of aspirin in breast cancer etiology.
Conclusions and Future Directions
The existing literature on the use of common over-the-counter and prescription medications has not definitively linked any of the drugs covered in this review to either increased or decreased risk of breast cancer. Important contributing factors to this apparent inconsistency are likely the numerous methodologic issues, discussed throughout this review, associated with the various study designs employed in these investigations. Thus, in conclusion, there is inconclusive evidence on the association between antibiotic use and breast cancer risk, no strong evidence pointing to a significant role of antidepressant and statin drugs in breast cancer development, somewhat inconclusive evidence on the effect of antihypertensive drugs, and significant chemoprotective evidence implicating aspirin use against breast cancer. Future studies with detailed lifetime medication histories are needed to further clarify these important associations. It is unlikely that such an assessment can be accomplished with a cohort study design, where repeated detailed medication measurement would be difficult to achieve. Thus, future case-control studies should consider in their design strategies for obtaining detailed and valid lifetime medication histories, which will likely involve a combination between self-report and prescription and/or health-care plan data. Further, in light of the strong and largely consistent findings from epidemiologic studies that link prolonged higher-dose aspirin use to reduce risk of breast cancer, a chemoprevention trial of NSAIDs or COX-2 inhibitors with similar chemopreventive properties to aspirin but without severe adverse gastrointestinal effects might be warranted. As pointed out above, medication use constitutes a ubiquitous exposure in the United States and in many countries worldwide. Given that breast cancer is the most common cancer in the United States and elsewhere, it is essential that we increase our understanding on the role of these commonly used drugs in the etiology of this disease.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Susan G. Komen for the Cure as part of the Environmental Factors and Breast Cancer Science Review project led by Silent Spring Institute with collaborating investigators at Harvard Medical School, Roswell Park Cancer Institute, and University of Southern California.