Interpretation and comparison with other studies
Statins decrease levels of serum cholesterol by inhibiting hydroxymethylglutaryl-coenzyme A reductase (HMG-CoAR), an enzyme involved in the biosynthesis of mevalonate acid in the mevalonate pathway of cholesterol synthesis [
9]. Statins have been suggested to exert anti-cancer effect through different pathways resulting in inhibition of cellular proliferation, induction of apoptosis, and suppression of tumour cell migration. Randomised clinical trials investigating the prognostic effect of statins in BC patients is lacking. A Danish phase III trial (Clinicaltrials.gov identifier: NCT04601116) started in 2021 and aims to recruit 3360 women with ER + BC, with the objective of evaluating the effect of statins on breast cancer prognosis (with invasive disease-free survival as the primary outcome measure). A few epidemiological studies have investigated the association between post-diagnostic use of statins and BC-specific survival, with inconclusive results [
25‐
29]. Consistent with our findings, some epidemiological studies report an association between post-diagnostic use of statins and a decreased risk of BC-specific death [
25,
27,
28], while others report no such association [
26,
28]. Based on registry data on 15,140 Scottish patients with BC, Mc Menamin et al. [
26] reported no association between post-diagnostic use of statins and risk of BC-specific death (HR = 0.93, 95% CI 0.77–1.12); however, an association with pre-diagnostic use of statins (HR = 0.85, 95% CI 0.74–0.98) was reported. Nowakowska et al. [
28] included 23,192 patients with BC identified in the Texas Cancer Registry and reported an association between post-diagnostic use of statins and a decreased risk of BC-specific death for patients with TNBC (HR = 0.42, 95% CI 0.20–0.88), but not for patients with non-TNBC (HR = 0.97, 95% CI 0.71–1.39). This is consistent with our findings in TNBC and non-TNBC patients.
The biological reason behind the association between post-diagnostic use of statins and a decreased risk of BC-specific death in patients with TNBC but not in patients with other molecular subtypes remains unclear. However, patients with TNBC receive chemotherapy more often than patients with other types of BC, and pre-clinical studies have suggested that statins exert a therapeutic effect through enhancing the effect of chemotherapeutic agents [
50,
51].
To corroborate the hypothesis of a potential interaction between statins and chemotherapeutic agents, we estimated a decreased risk of BC-specific death associated with use of statins among recipients of chemotherapy (HR = 0.79, 95% CI 0.63–0.98) but not among patients who do not receive chemotherapy (HR = 1.01, 95% CI 0.82–1.24) (Fig.
2).
Metformin functions by reducing resistance to insulin and decreasing serum levels of insulin [
10]. Pre-clinical studies have suggested that metformin inhibits cancer progression and prognosis via direct effects on the cancer cells, by acting on the AMP-activated protein kinase (AMPK)/mammalian target of rapamycin (mTOR) pathway, and indirect effects by decreasing serum levels of insulin and insulin-like growth factor 1 (IGF-1). The association between post-diagnostic use of metformin and BC-specific survival has been studied in a small number of epidemiological studies [
31,
32,
34]. Our finding of a decreased risk of BC-specific death associated with post-diagnostic use of metformin corroborates both pre-clinical studies and previous epidemiological studies evaluating the association in patients with both BC and diabetes mellitus type II [
10,
31,
32,
34]. Kim et al. [
34] evaluated 386 South Korean diabetic patients with BC and reported a decreased risk of BC-specific death associated with post-diagnostic use of metformin, compared to non-metformin antidiabetics, in patients with ER + and/or PR + BC but not in patients with BC with both ER- and PR-. The association with a decreased risk of BC-specific death among metformin users with ER +/PR + BC but not with ER − and PR − BC is consistent with our finding of an association in patients with ER + BC only. Furthermore, it corroborates the hypothesis that the AMPK/mTOR pathway plays a role in the development of resistance to endocrine therapy in ER + BC and that the metformin activity on the AMPK/mTOR pathway can re-sensitise the ER + BCs to endocrine therapy [
52]. In conflict with our findings, a randomised clinical trial published in 2022 by Goodwin et al. [
13], including 3649 BC patients without diabetes, reported that addition of metformin to standard treatment did not improve invasive disease-free survival. The estimate did not differ by ER status. The Goodwin trial confirmed the findings from previously published smaller randomised clinical trials [
12].
There are plausible mechanisms suggesting that aspirin affect BC progression and prognosis by altering levels of prostaglandins [
8]. Aspirin inhibits cyclooxygenase (COX), an enzyme involved in the biosynthesis of prostaglandins, which are mediators of inflammation and pain. Prostaglandins are suggested to promote cellular proliferation and invasiveness, and stimulate the activity of aromatase, an enzyme responsible for the biosynthesis of oestrogens, which are drivers of ER +/luminal BC. In addition, prostaglandins are the precursors of thromboxane, which is required to facilitate platelet aggregation, and the antiplatelet effect of aspirin has been suggested to inhibit tumour cells from initiating metastases. The results reported by epidemiological studies assessing the prognostic effect of the post-diagnostic use of aspirin (including both low-dose and regular dose) are inconclusive [
14‐
20]. Our findings of no association corroborate a number of previous epidemiological studies [
17‐
20], as well as the recently finished randomised phase III Aspirin after Breast Cancer (ABC) trial [
11], which included 3021 BC patients and reported that addition of aspirin (300 mg) to the standard treatment did not improve disease-free survival. In contrast, some epidemiological studies have reported an association with longer BC-specific survival [
14‐
16]. Using data from the Iowa Women’s Health Study (591 patients with BC), Blair et al. [
14] reported that post-diagnostic use of aspirin was associated with a decreased risk of BC-specific death (HR = 0.53, 95% CI 0.30–0.93). Similarly, Holmes et al. [
15] reported that use of aspirin after diagnosis was associated with a decreased risk of BC-specific death (2–5 days a week: HR = 0.29, 95% CI 0.16–0.52; 6–7 days a week: HR = 0.36, 95% CI 0.24–0.54) in 4164 patients with BC participating in the Nurses’ Health Study. The results did not differ when stratified by stage, menopausal status, body mass index, or ER-receptor status. In addition, a Scottish registry study by Fraser et al. (4627 patients with BC) [
16] reported an association between post-diagnostic use of aspirin and a decreased risk of BC-specific death (HR = 0.53, 95% CI 0.45–0.63). The strength or dose of aspirin may have been inconsistent between studies. Similar to our study, all the epidemiological studies that reported no association included almost exclusively low dose-aspirin users [
17‐
20], while neither the Iowa Women’s Health Study nor the Nurses’ Health Study restricted their questionnaires to users of low-dose aspirin, nor did they collect information on the dose of aspirin for the surveys used in the studies by Blair et al. and Holmes et al. [
14,
15]. Aspirin in higher doses is rarely used in Norway [
53], but more frequently used in the USA [
54]. Therefore, it is possible that the Iowa Women’s Health Study and the Nurses’ Health Study included a non-neglectable proportion of users of aspirin in higher doses. One hypothesis is that high doses of aspirin are necessary to see an effect on BC prognosis.
Breast cancer treatments, such as chemotherapy, have well-documented cardio-toxic side effects, potentially leading to increased risk of all-cause deaths (driven by cardiovascular related deaths). Giving medications frequently used to prevent or treat cardiovascular diseases in addition to the chemotherapy might prevent these side effects, resulting in prolonged overall survival. Previous studies assessed the association for statins, metformin, and aspirin with overall survival in patients with BC, have been conflicting, reporting no association and associations with both longer and shorter overall survival [
14,
16,
18,
19,
25‐
28,
31,
32]. However, most of the previous studies, in line with our study, have reported that post-diagnostic use of statins and metformin were associated with a decreased risk of all-cause death. In contrast with most studies, we found that use of low-dose aspirin was associated with an increased risk of all-cause death, potentially due to an increased risk of cardiovascular-related deaths among the users of low-dose aspirin.
Strengths and limitations
The main strength of our cohort study is the population-based design with data from nationwide registries of high quality and completeness, minimising the risk of misclassification bias and selection bias. The use of a prescription database avoided self-reported use of drugs, which may be less accurate and are associated with a higher risk of introducing misclassification bias. Another strength was the large sample size and the detailed information on tumour characteristics that allowed for exploring the potential prognostic effect in depth.
However, there are several limitations in our study that needs mentioning. First, the Norwegian Prescription Database records information on filled prescriptions but it contains no information on actual use or adherence, and the database only includes information on dispensed medications from community pharmacies and not medications given at hospitals or nursing homes. This may have resulted in some misclassification of both users and non-users, possibly leading to underestimation of the associations between use of the medications and survival (HR biased towards 1). Second, patients contributed person time to the user group of a specific medication from the date they fulfilled the criteria of post-diagnostic use until end of follow-up. This definition of exposure might not capture the real time-dependent exposure. By handling medication use in this way we presumed that the potential prognostic effect would last after discontinuation for patients who discontinued the medication before end of follow-up. If this presumption does not hold, then the estimated difference in survival between users and no users would be smaller than the true difference between the groups, and the estimated associations between post-diagnostic use and survival would have been underestimated (HR biased towards 1). Third, we did not have access to information on comorbid conditions. However, this was addressed by using the information on dispensed medications as proxy for comorbid conditions. Fourth, considering the high proportion of patients with missing information on chemotherapy use, the results stratified by chemotherapy use should be interpreted with caution. Fifth, that we did not use active comparators for low-dose aspirin and statins may have introduced bias through confounding by indication. However, the fact that the conditions treated with low-dose aspirin and statins are not clearly associated with the risk of BC-related death may somewhat alleviate this concern. Sixth, we missed information on important confounders, such as the body mass index. This may have biased the results. For example, if high body mass index is associated with increased use of the medications studied and an increased risk of BC-related death, not including body mass index possibly resulted in a bias towards an increased HR. Finally, some of the subgroups have few events, making it difficult to draw conclusions form these subgroups.