Statin Prescriptions and Breast Cancer Recurrence Risk: A Danish Nationwide Prospective Cohort Study

HMG-CoA reductase inhibitors (statins) block the rate-limiting step in cholesterol biosynthesis (1). These drugs have become standard therapy to manage hypercholesterolemia and associated morbidities (2). Additionally, statins may affect the occurrence or outcomes of other diseases—including cancer—either by downstream consequences of cholesterol reduction or by mechanisms outside of the cholesterol synthesis pathway (3,4). Such pleiotropic effects may be limited to the lipophilic statins (5,6). Between 1994 and 2008, the prevalence of statin use among people age 30 years or older in northern Denmark increased from 1.1% to 36% (7), a pattern similar to that observed in the United States (8). Breast cancer remains the most common malignancy among women, accounting for 28% of all cancer diagnoses in American women (9). Any effect of statin use on breast cancer outcomes would therefore have important public health and clinical implications. Earlier studies have examined the association between statin use and breast cancer incidence. Most studies have reported no association between breast cancer incidence and statin use (10–15), although some reports suggested inverse associations may exist (16,17). Several meta-analyses have provided data that suggest that no relationship exists between statin use and breast cancer incidence (18–24). Only one study has examined the association between postdiagnosis statin use and breast cancer recurrence. Kwan et al. (25) followed a cohort of approximately 2000 breast cancer survivors for a mean of 5 years and observed an association between lipophilic ARTICLE

HMG-CoA reductase inhibitors (statins) block the rate-limiting step in cholesterol biosynthesis (1). These drugs have become standard therapy to manage hypercholesterolemia and associated morbidities (2). Additionally, statins may affect the occurrence or outcomes of other diseases-including cancer-either by downstream consequences of cholesterol reduction or by mechanisms outside of the cholesterol synthesis pathway (3,4). Such pleiotropic effects may be limited to the lipophilic statins (5,6).
Between 1994 and 2008, the prevalence of statin use among people age 30 years or older in northern Denmark increased from 1.1% to 36% (7), a pattern similar to that observed in the United States (8). Breast cancer remains the most common malignancy among women, accounting for 28% of all cancer diagnoses in American women (9). Any effect of statin use on breast cancer outcomes would therefore have important public health and clinical implications. Earlier studies have examined the association between statin use and breast cancer incidence. Most studies have reported no association between breast cancer incidence and statin use (10)(11)(12)(13)(14)(15), although some reports suggested inverse associations may exist (16,17). Several meta-analyses have provided data that suggest that no relationship exists between statin use and breast cancer incidence (18)(19)(20)(21)(22)(23)(24).
Only one study has examined the association between postdiagnosis statin use and breast cancer recurrence. Kwan et al. (25) followed a cohort of approximately 2000 breast cancer survivors for a mean of 5 years and observed an association between lipophilic

ARTICLE Statin Prescriptions and Breast Cancer Recurrence Risk: A Danish Nationwide Prospective Cohort Study
Background Accumulating evidence suggests that statins affect diseases other than cardiovascular disease, including cancer, and that these effects may depend on the lipid solubility of specific statins. Though many studies have reported an association between statin use and breast cancer incidence, the relationship between statin use and breast cancer recurrence has not been well studied.

Methods
We conducted a nationwide, population-based prospective cohort study of all female residents in Denmark diagnosed with stage I-III invasive breast carcinoma who were reported to the Danish Breast Cancer Cooperative Group registry between 1996 and 2003 (n = 18 769). Women were followed for a median of 6.8 years after diagnosis. Prescriptions for lipophilic and hydrophilic statins were ascertained from the national electronic pharmacy database. Associations between statin prescriptions and breast cancer recurrence were estimated with generalized linear models and Cox proportional hazards regression with adjustment for age and menopausal status at diagnosis; histological grade; estrogen receptor status; receipt of adjuvant therapy; type of primary surgery received; pre-diagnosis hormone replacement therapy; and co-prescriptions of aspirin, angiotensin-converting enzyme inhibitors, nonsteroidal anti-inflammatory drugs, or anticoagulants. All statistical tests were two-sided.

Results
Most prescriptions for lipophilic statins in the study population were for simvastatin. Exclusive simvastatin users experienced approximately 10 fewer breast cancer recurrences per 100 women after 10 years of follow-up (adjusted 10-year risk difference = 20.10, 95% confidence interval = 20.11 to 20.08), compared with women who were not prescribed a statin. Exclusive hydrophilic statin users had approximately the same risk of breast cancer recurrence as women not prescribed a statin over follow-up (adjusted 10-year risk difference = 0.05, 95% confidence interval = 20.01 to 0.11).
statin use and decreased cancer recurrence. However, the estimate was not measured precisely enough to provide strong evidence against a null association (hazard ratio [HR] = 0.67, 95% confidence interval [CI] = 0.39 to 1.13) (25). An accompanying editorial (26) hypothesized that functional similarities between statins and bisphosphonates might yield an inverse association between lipophilic statin use and bone recurrences in particular. Others have hypothesized that any association between statins and recurrence could depend on tumor estrogen receptor (ER) status (27)(28)(29) or histological grade (28). Finally, the preventative activity of statins against cancer recurrence by the potentiation of radiotherapy has also been described (30)(31)(32).
The aims of this study were to measure the association between statin use and the recurrence rate of breast cancer in a large population-based cohort of breast cancer survivors, to examine whether the association differs by drug solubility or specific recurrence site, and to evaluate modification of the association by ER status, histological grade, and receipt of radiotherapy.

Subjects and Methods
This study was approved by the Danish Data Protection Agency (journal number 2005-41-5396).

Source Population and Data Collection
We enrolled all female residents of Denmark who were diagnosed with incident invasive breast carcinoma between 1996 and 2003, and who were reported to the Danish Breast Cancer Cooperative Group (DBCG) registry. Beginning in 1977, the DBCG has registered most invasive breast cancers diagnosed in Danish women (33,34) and completeness of breast cancer registration by the DBCG ranged from 87%-96% for the period 1986-1997 (35). Data on tumor, treatment, and patient characteristics were collected by the DBCG from treating physicians using standardized forms. Follow-up examinations were done every 3-6 months for the first 5 years post-diagnosis and then annually for years 6-10 (34). Follow-up assessments included a physical examination and, if indicated, a chest x-ray, bone scan, or other investigations to detect recurrence. Women presenting with a recurrence between examinations were also reported to the registry.
To ascertain prescription drug exposures, we used each patient's unique Civil Personal Registry number to link her DBCG data to the National Registry of Medicinal Products (NRMP) maintained by Statistics Denmark (Copenhagen, Denmark). The NRMP has automatically recorded all prescriptions dispensed at Danish pharmacies, beginning from 1995. Each record logs the patient's Civil Personal Registry number, the medication dispensed [classified by the Anatomic Therapeutic Chemical system (36)], the date, and quantity dispensed. Medical history variables were ascertained by linking the cohort to the Danish National Registry of Patients, which has recorded diagnoses made at nonpsychiatric hospitals, beginning from 1977.

Definitions of Analytic Variables
The age at diagnosis was categorized into decades for stratified analyses but was used as a continuous variable in regression models. We defined each patient's follow-up time as the number of days between breast cancer surgery and the diagnosis of recurrent breast cancer, death, emigration from Denmark, accrual of 10 years of follow-up, or December 31, 2008, whichever came first. Recurrent breast cancer is defined by DBCG protocol as any local, regional or distant recurrence, or contralateral breast cancer. We examined whether the association between statin use and recurrence differed by the following anatomical sites: bone, lymph nodes, ipsilateral breast, contralateral breast, lung, liver, or the central nervous system.
We classified statin exposure as a time-varying dichotomous variable that was updated yearly after breast cancer surgery. In each yearly interval, patients were classified as exposed to statins if they had at least one prescription logged in the NRMP with an Anatomic Therapeutic Chemical code beginning with "C10AA." Patients prescribed a statin were assumed to be exposed (ie, statin users), whereas patients who were not prescribed a statin were assumed to be unexposed to statins (ie, nonusers) and will be referred to as such from here on out. Statin exposure was further characterized by solubility (hydrophilic or lipophilic) and by specific drugs (5,37). For solubility-based exposure definitions, we identified women who were exclusively prescribed lipophilic statins, hydrophilic statins, or simvastatin during the follow-up. Women who switched drug categories during follow-up were excluded from analyses using these definitions (n = 552). To evaluate

Prior knowledge
Statins, a class of drugs that block the cholesterol synthesis, may affect the occurrence or outcomes of diseases other than cardiac disease, including cancer. Although most reports suggest that there is no association between breast cancer incidence and statin use, there are few reports investigating the relationship between breast cancer recurrence and statin use.

Study design
Associations between statin use and breast cancer recurrence among a cohort of Danish women diagnosed with invasive breast cancer were assessed. Records of all statin prescriptions were available for each woman so that statin users were defined as those who had been prescribed a statin, whereas nonusers were defined as those who were not prescribed a statin. Patients were followed for a median of 6.8 years to allow time to assess breast cancer recurrences.

Contribution
Patients using simvastatin, a lipophilic statin, experienced fewer breast cancer recurrences compared with nonusers. Hydrophilic statin use was not associated with breast cancer recurrence.

Implication
Simvastatin use is associated with a decreased risk of breast cancer recurrence. Other lipophilic statins should be studied for similar associations and mechanistic studies to determine the underlying biology behind the observations reported here should be performed.

Limitations
Having been given a statin prescription was assumed to be the equivalent of having taken a statin. The rate of compliance in the study cohort is unknown.
From the Editors the duration and intensity of statin use, we tabulated each patient's total number of years of exposure and the total number of statin prescriptions filled.
We also ascertained the exposure to comedications that could potentially confound the association between statin use and breast cancer recurrence. Exposures to aspirin (38), nonsteroidal anti-inflammatory drugs (39,40), angiotensin-converting enzyme inhibitors (41), and vitamin K anticoagulants (42) during follow-up were identified by Anatomic Therapeutic Chemical codes starting with "B01AC06," "M01A," "C09," and "B01AA," respectively. Pre-diagnosis combination hormone therapy (43) was defined by Anatomic Therapeutic Chemical codes starting with "G03F." We summarized each patient's medical history from 1977 until her breast cancer diagnosis by searching the Danish National Registry of Patients for the diagnoses that comprise the Charlson Comorbidity Index (44). Tumor stage was classified according to Union for International Cancer Control guidelines and summarized as stage I, II, or III. Histological grade was classified as low, moderate, high, or missing if there was insufficient or unsuitable tissue for determination. Tumor ER status (positive or negative) and receipt of adjuvant endocrine therapy (ET; tamoxifen or aromatase inhibitor) were summarized into a four-category variable for use in regression models (ER+/ET+, ER2/ET2, ER+/ET2, ER2/ET+). Receipt of adjuvant chemotherapy and radiotherapy were defined dichotomously. Menopausal status (pre-or postmenopausal) was defined on the breast cancer diagnosis date.

Statistical Methods
We tabulated the frequency and proportion of patients, recurrences, and person-years according to patient, tumor, and treatment characteristics, and exposure to statins and other medications. We estimated the 5-and 10-year recurrence hazard ratios and 95% confidence intervals between statin groups in unadjusted and multivariable Cox regression models, with statin and other drug exposures characterized as time-varying covariables lagged by 1 year. These lagged models impose a reasonable induction period for an effect of statins and co-prescriptions on recurrence and guard against the possibility that imminent recurrences affected prescribing patterns. We stratified lagged models by ER status, histological grade, and type of primary therapy (mastectomy without radiotherapy or breast-conserving surgery with radiotherapy) to evaluate modification of the association by these characteristics. We used competing risks Cox regression, in which recurrences at sites other than the site of interest were censored, to estimate 10-year hazard ratios for anatomical site-specific recurrence (26,45). Proportionality of hazard functions for statin exposure definitions (exclusive hydrophilic, lipophilic, or simvastatin use) were verified by checking that Wald tests for the interaction between statin exposures and the logarithm of person-time were non-statistically significant (two-tailed a = 0.05) (45). Adjusted risks and risk differences were estimated with generalized linear regression models, using a normal probability distribution with robust variance estimation and an identity link function (46). All statistical analyses were performed with SAS version 9.2 (SAS Institute, Cary, NC). All statistical tests were two-sided (a = 0.05).

Baseline Characteristics of the Cohort
We included 18 769 female residents of Denmark diagnosed with primary invasive breast carcinoma between 1996 and 2003. There were 3419 breast cancer recurrences recorded over 114 006 person-years of follow-up (median = 6.8 years). Table 1 shows the distribution of patients according to statin exposure and key demographic, tumor, and treatment variables. During follow-up, users of any statin were older, were more likely to be postmenopausal at the time of breast cancer diagnosis, and had more comorbid conditions compared with nonusers. Furthermore, statin users were more likely to have ER-positive tumors and to receive adjuvant endocrine therapy but were less likely to receive adjuvant chemotherapy. Compared with nonusers, statin users were also more likely to have pre-diagnosis prescriptions for combination hormone therapy (

Statin Prescriptions in the Cohort
There were 3282 breast cancer patients who were ever prescribed a statin (statin users) during follow-up. Of these, 2524 patients were prescribed lipophilic statins exclusively and 206 were prescribed hydrophilic statins exclusively. Simvastatin, the most lipophilic statin, accounted for 92% of lipophilic statin prescriptions in the cohort. The median duration of exposure was 4 years (interquartile range = 2-6 years), and the median number of prescriptions filled was 11 (interquartile range = 5-20 prescriptions). Table 2 shows the number of specific statins (simvastatin, lovastatin, fluvastatin, cerivastatin, atorvastatin, pravastatin, and rosuvastatin) prescribed during the follow-up and their log partition coefficients (higher values of which correspond to greater lipophilicity). Table 3 summarizes the estimated associations between statin exposure and 5-and 10-year breast cancer recurrence. During a maximum of 10 years of follow-up, there were 249 breast cancer recurrences among statin users (adjusted recurrence risk = 0.207). Of these, 182 recurrences occurred among lipophilic statin users who were exclusively prescribed simvastatin, lovastatin, fluvastatin, or cerivastatin (adjusted recurrence risk = 0.194), and 39 occurred among hydrophilic statin users who were exclusively prescribed atorvastatin, pravastatin, or rosuvastatin (adjusted recurrence risk = 0.350). Among nonusers, there were 3170 recurrences (adjusted recurrence risk = 0.302).

Associations between Lipophilic and Hydrophilic Statin Use and Breast Cancer Recurrence
Exclusive hydrophilic statin users had approximately the same rate of breast cancer recurrence as nonusers (10-year risk difference = 0.05, 95% CI = 20.01 to 0.11; 10-year adjusted HR = 1.2, 95% CI = 0.79 to 1.7), whereas exclusive lipophilic statin users had a reduced rate of recurrence compared with nonusers (10-year adjusted HR = 0.73, 95% CI = 0.60 to 0.89) ( Table 3). When we examined specific drugs within the lipophilic class, we found an inverse association between exclusive simvastatin use and the rate of recurrence (10-year adjusted HR = 0.70, 95% CI = 0.57 to 0.86), and exclusive simvastatin users experienced approximately 10 fewer breast cancer recurrences per 100 women after 10 years of follow-up (adjusted 10-year risk difference = 20.10, 95% CI = 20.11 to 20.08) compared with women who were not prescribed a statin. However, low outcome frequency precluded the modeling of recurrence associations for exclusive use of the other statins (lovastatin, fluvastatin, and cerivastatin). We therefore focused our lipophilic analyses on exclusive simvastatin users. The lack of an association between hydrophilic statin use and breast cancer recurrence presented an opportunity to address potential confounding by indication, whereby medical indications for statin therapy might themselves modify recurrence risk. We conducted a separate analysis using exclusive hydrophilic statin users as a reference group for exclusive simvastatin users. To our knowledge, the prescription of a statin was not influenced by the solubility of different statins during the study period, and so lipophilic and hydrophilic statin users should have similar distributions of medical indications. We observed minor differences in the distribution of tumor stage and ER expression/receipt of ET between exclusive simvastatin and hydrophilic statin users, but these were included as covariables in adjusted statistical models. Exclusive simvastatin users again had a reduced rate of breast cancer recurrence when compared with exclusive hydrophilic statin users (10-year adjusted HR = 0.55, 95% CI = 0.35 to 0.85) ( Table 3). * CI = confidence interval; ER = estrogen receptor; ET = endocrine therapy; HR = hazard ratio. † All prescription exposure characterizations were updated yearly over follow-up and coded as time-varying variables. All prescription exposures were lagged by 1 year. ‡ Adjusted for age at diagnosis (continuous), menopausal status at diagnosis (pre-or postmenopausal), International Union Against Cancer stage (I, II, or III), histological grade (low, moderate, or high), ER status and receipt of adjuvant ET (ER2/ET2, ER2/ET+, ER+/ET2, ER+/ET+), receipt of adjuvant chemotherapy (yes or no), type of primary surgery received (mastectomy, breast-conserving surgery with radiotherapy), pre-diagnosis exposure to combination hormone replacement therapy (yes or no), and co-prescriptions (time varying, updated yearly) of aspirin, angiotensin-converting enzyme inhibitors, nonsteroidal anti-inflammatory drugs, or anticoagulants (yes or no). § Estimated with Cox proportional hazards regression models; 95% CIs were calculated by the Wald method.
|| Estimated with generalized linear regression models, using a normal probability distribution, robust variance estimation, and an identity link function. 95% CIs were calculated by the Wald method. Associations between statin use and the rate of breast cancer recurrence were essentially unchanged when we restricted the analysis to women who did not receive chemotherapy (10-

Simvastatin Use and Site of Breast Cancer Recurrence
The total number of anatomical site-specific recurrences for exclusive simvastatin users and nonusers combined, and the 10-year recurrence hazard ratio at each site, comparing exclusive simvastatin users with nonusers, are given in Table 4. The majority of events were cancer recurrences in the bone (1087 of 3336 events, 33%), but an association with simvastatin use was not found (adjusted HR = 0.91, 95% CI = 0.65 to 1.3). Recurrences in the ipsilateral breast (550 of 3336 events, 17%) appeared to be associated with simvastatin use (adjusted HR = 0.57, 95% CI = 0.32 to 1.0), as did recurrences in the contralateral breast (520 of 3336 events, 16%; adjusted HR = 0.54, 95% CI = 0.33 to 0.90) and lymph nodes (527 of 3336 events, 16%; adjusted HR = 0.56, 95% CI = 0.31 to 1.0). The remaining recurrence sites each accounted for less than 10% of the 3336 events, and their associations with simvastatin use were measured with poor precision. Also, stratified models showed inverse associations between simvastatin use and breast cancer recurrence in all categories of ER status, histological grade, and administration of radiotherapy.

Discussion
We followed 18 769 breast cancer survivors for a median of 6.8 years and observed a substantial decrease in the rate of recurrence among exclusive users of simvastatin, the most lipophilic of the statins prescribed to women in our cohort. There were few recurrences among exclusive users of lovastatin, fluvastatin, or cerivastatin, which precluded estimation of recurrence associations for these other lipophilic drugs. Hydrophilic statin use did not appear to be associated with breast cancer recurrence. These findings were robust to adjustment for established clinical risk factors for breast cancer recurrence, potentially confounding comedications, and potential confounding by indication.
The inverse association between simvastatin use and breast cancer recurrence that we observed is consistent with that reported by Kwan et al. (25) in a smaller study, in which statin users were mostly prescribed simvastatin and lovastatin. We observed similar associations between simvastatin use and many of the specific recurrence sites we evaluated. Simvastatin use was not associated with cancer recurrence at the bone, contrary to a report that an inverse association with statins may be limited to this site (26). We also did not observe substantial modification in the relationship by ER status, histological grade, or receipt of radiotherapy.
The chief strengths of this study are its large size, prospective design, and the use of high-quality prescription and clinical registry data. The DBCG registry combines the advantages of data validity and completeness of follow-up possible in a clinical trial with the advantages of studying a nearly complete and unselected source population.
There are also limitations in our study. Misclassification of analytic variables could have biased the observed associations, though the impact is likely small. The positive predictive value for classification of breast cancer recurrence by the DBCG registry equaled 99.4% in a medical record validation study (48). Furthermore, of 1888 recurrences identified in medical records among 4455 breast cancer patients assigned to DBCG treatment protocols, 1813 (96%) were correctly registered as recurrences in the database, 74 (3.9%) were identified as breast cancer deaths, and only one (0.05%) was not identified as either a recurrence or a breast cancer death. Another limitation to our study is that we could not confirm that women who were prescribed statins actually complied. However, records are logged in the registry only after a patient has presented their prescription to a pharmacist and paid for the dispensed medication. Thus, we expect that logged prescriptions are received by patients intending to comply. Statins are only available by prescription in Denmark, therefore patients classified as nonusers are assumed to be completely unexposed to statins. Also, given the prospective design, and because prescription and recurrence data were extracted from independent registries, differential misclassification of statin exposure or breast cancer recurrence are implausible explanations for our findings.
Another potential limitation of our study is that we could not adjust for body mass index because the data were unavailable. A previously published study in an overlapping population showed that body mass index was associated only with distant breast cancer recurrences (49). Because we found inverse associations for local, regional, and distant recurrences-and because body mass index is positively associated with both statin use and recurrence riskconfounding by body mass index is unlikely to account for our observations. Furthermore, we do not expect substantial residual confounding of our adjusted estimates. Adjustment for factors strongly associated with recurrence did not attenuate our associations, and another uncontrolled factor strongly related to both statin use and cancer recurrence is unlikely to exist.
Also, hypercholesterolemia is associated with impaired angiogenesis and a lower risk of recurrence among women with invasive ductal carcinoma (50), and if women with hypercholesterolemia, the major indication for statin therapy, have an inherently lower risk of breast cancer recurrence, then confounding by indication could explain our results (51). Because the medical indications should be identical for lipophilic and hydrophilic statins, and because we observed an inverse association between lipophilic statin use and recurrence, such a bias is unlikely. Furthermore, when we directly compared lipophilic statin users with hydrophilic statin users in an effort to address confounding by indication, we obtained a very similar association with recurrence when lipophilic statin users were compared with nonusers of statins.
The inverse association we observed between simvastatin use and breast cancer recurrence may seem discordant with metaanalyses that reported no association between statin use and breast cancer incidence (18)(19)(20)(21)(22)(23)(24). It is possible that some past incidence studies did not observe an association because a high proportion of statin users took hydrophilic drugs. For instance, Cauley et al. (16) observed a lower risk of breast cancer among lipophilic statin users, 82% of whom took either simvastatin or lovastatin (HR = 0.82, 95% CI = 0.70 to 0.97); and Boudreau et al. (17) observed an inverse association with overall statin use, when approximately 48% of the study participants were simvastatin or lovastatin users (odds ratio = 0.7, 95% CI = 0.4 to 1.0) (17). However, Woditschka et al. (14) observed no association between statin use and breast cancer incidence in which most users were exposed to the lipophilic drugs simvastatin and lovastatin (odds ratio = 1.02, 95% CI = 0.97 to 1.08) (52); and Friis et al. (12) studied statin use and breast cancer incidence in a Danish population that overlaps with the population analyzed in our study (so simvastatin is expected to have accounted for a large proportion of the overall statin exposure) and also observed no association; (relative risk = 1.02, 95% CI = 0.76 to 1.36) (12). Although the apparent discordance between the statin associations with breast cancer incidence and recurrence may be explained by inadequate exposure characterization with respect to solubility, it is also important to realize that factors which prevent recurrence are not necessarily expected to also prevent incidence.