Aspirin Use and Survival Among Patients With Breast Cancer: A Systematic Review and Meta-Analysis

Abstract Background Previous meta-analyses have indicated that aspirin could affect breast cancer outcomes, particularly when taken post-diagnostically. However, several recent studies appear to show little to no association between aspirin use and breast cancer mortality, all-cause mortality, or recurrence. Aims This study aims to conduct an updated systematic review and meta-analysis on the associations of pre-diagnostic and post-diagnostic aspirin use with the aforementioned breast cancer outcomes. It also looks, through subgroup analyses and meta-regressions, at a range of variables that could explain the associations between aspirin use and breast cancer outcomes. Results In total, 24 papers and 149 860 patients with breast cancer were included. Pre-diagnostic aspirin use was not associated with breast-cancer-specific mortality (HR 0.98, 95% CI, 0.80-1.20, P = .84) or recurrence (HR 0.94, 95% CI, 0.88-1.02, P = .13). Pre-diagnostic aspirin was associated with non-significantly higher all-cause mortality (HR 1.27, 95% CI, 0.95-1.72, P = .11). Post-diagnostic aspirin was not significantly associated with all-cause mortality (HR 0.87, 95% CI, 0.71-1.07, P = .18) or recurrence (HR 0.89, 95% CI, 0.67-1.16, P = .38). Post-diagnostic aspirin use was significantly associated with lower breast-cancer-specific mortality (HR 0.79, 95% CI, 0.64-0.98, P = .032). Conclusions The only significant association of aspirin with breast cancer outcomes is lower breast-cancer-specific mortality in patients who used aspirin post-diagnostically. However, factors such as selection bias and high inter-study heterogeneity mean that this result should not be treated as conclusive, and more substantial evidence such as that provided by RCTs is needed before any decisions on new clinical uses for aspirin should be made.


Introduction
In 2020, there were an estimated 2.26 million new cases of female breast cancer and around 685 000 female breast cancer deaths worldwide. 1These figures make breast cancer both the most prevalent cancer in women and the cancer responsible for the greatest number of deaths in women.Finding new treatments for such a large contributor to global morbidity and mortality is a hugely desirable prospect.
Aspirin is a commonly used, inexpensive NSAID.Preclinical studies suggest a range of plausible mechanisms by which the effects of aspirin could improve survival in patients with breast cancer.One example is the inhibition of COX-2, overexpression of which has been associated with lower overall survival in patients with breast cancer. 2Another is the inhibition of platelet function, given that platelets are thought to contribute to cancer metastasis. 3he ADD-Aspirin trial 4 is currently investigating the post-diagnostic use of aspirin in the treatment of breast cancer, but its primary completion date is estimated to be October 2026. 5A meta-analysis of observational evidence could therefore be useful until this point in assessing the potential value of aspirin as a treatment for breast cancer.
Multiple prior meta-analyses [6][7][8][9] have found that aspirin, especially when taken after diagnosis of breast cancer, may be associated with reduced breast-cancer-specific mortality.However, 2 of these earlier meta-analyses were carried out a significant time ago, with literature searches up to 2014, 6,7 whilst the more recent 2 did not distinguish between prediagnostic and post-diagnostic aspirin use. 8,9This study aims e2 The Oncologist, 2024, Vol. 29, No. 1   to build on these meta-analyses and conduct an updated and more detailed investigation into the associations between aspirin and mortality and recurrence in patients with breast cancer.

Search
We searched Ovid MEDLINE (including records from 1946 onwards) and Ovid EMBASE (including records from 1974 onwards), with publications up to and including June 30th, 2021 included.The full search strategy is shown in Appendices A and B. In addition to this, the reference lists of papers included in the secondary screening process were checked for relevant studies.After the primary screening process was completed, any papers identified as potentially relevant were screened independently by both authors, with differences resolved by discussion.

Inclusion Criteria
RCTs, cohort studies, and case-control studies were eligible for inclusion, while reviews, meta-analyses and conference or poster abstracts with insufficient data were ineligible.Papers containing risk estimates and 95% CIs for breast-cancer-specific mortality, all-cause mortality, or recurrence were eligible for inclusion, as were those with sufficient data to calculate risk estimates.Only papers containing data on aspirin (and not NSAIDs as a whole or other NSAIDs) as the primary exposure were included.

Data Extraction and Quality Assessment
AB extracted data from the selected papers, these data were then reviewed by CK.The categories of data extracted are given in Appendix C. We then performed a quality assessment using the nine-point Newcastle-Ottawa scale 10 (NOS); studies scoring seven points or higher were considered high quality.

Statistical Analysis
Analyses were carried out using R statistical software. 11,12hen analyzing risk estimates, RRs and ORs were considered to be approximations of HRs, and all 3 measures were used interchangeably in common and random effects meta-analyses.Where ORs, RRs, or HRs were not given in the texts of studies, available data was used to calculate ORs or RRs.Where possible, RRs, ORs, or HRs which had been adjusted for confounders were used in meta-analyses, with unadjusted risk estimates only being included in analyses when papers did not provide adjusted estimates.Interstudy heterogeneity was assessed using the I 2 statistic, 13 with the heterogeneity of ≥50% considered substantial. 14ensitivity analyses were used to assess sources of heterogeneity.Publication bias was evaluated using Egger's test. 15eta-regressions 16 were used to assess the effect of individual variables on risk estimates.These variables were: year of publication, sample size, cancer stage at diagnosis, cancer grade at diagnosis, mean/median age, menopause status, smoking status, tumor molecular subtypes, cancer treatment, race, node-negative, or positive cancer at diagnosis, BMI, mean/median follow-up length, and the proportion of patients on low-dose or high-dose aspirin in a study (Supplementary Table S1).

Characteristics of Included studies
A flowchart summarizing the search and study selection process is shown in Fig. 1.A total of 1325 papers identified through database searching were screened by title and abstract, with 1220 excluded due to irrelevance to this study, being a review or meta-analysis, or not having an available full-text version.A total of 105 studies had their full texts screened; 82 were excluded, with the majority of exclusions due to the paper's primary outcome being breast cancer risk rather than an outcome relevant to this study.Screening the reference lists of these included studies yielded a further 7 potentially relevant studies, 1 17 of which were eventually included in our meta-analysis.This meant a total of 24  papers were chosen for the meta-analysis; 21 cohort studies  and 3 case-control studies [38][39][40] were included. Key cracteristics of these papers are shown in Tables 1 and 2; confounders that they used to calculate adjusted risk estimates are listed in Supplementary Table S2.Altogether, 149 860 patients with breast cancer were enrolled in the studies in this metaanalysis.Supplementary Tables S3 and S4 show the methodological quality of the included studies; the mean NOS score was 7.8, indicating that included studies were generally of high quality.

Pre-Diagnostic Aspirin Use and Breast-Cancer-SpecificMortality
There is no significant association of pre-diagnostic aspirin use with breast-cancer-specific mortality in either common (HR 1.03, 95% CI, 0.95-1.11,P = .47)or random (HR 0.98, 95% CI, 0.80-1.20,P = .84)effects models (Fig. 2a), as well as a large amount of heterogeneity (I 2 = 85%).However, sensitivity analysis involving the removal of single papers to find sources of heterogeneity showed that the removal of Fraser et al. 23 led to significantly decreased heterogeneity (I 2 = 0%, 95% CI, 0%-62.4%).The risk estimate for breast-cancer-specific mortality also becomes significant when Fraser et al.'s paper is removed, albeit only barely (HR 0.91, 95%, CI 0.84-0.99,P = .022).Meta-analysis of additional data contained within 3 studies 27,31,32 showed that patients who took less than 1 dose of aspirin per day had a similar likelihood of dying of breast cancer as those who took more than 1 dose of aspirin per day (HR = 0.90, 95% CI, 0.75-1.07,P = .24for <1 daily dose and HR = 0.95, 95% CI, 0.75-1.19,P = .65for >1 daily dose) (Supplementary Fig. S1).No evidence of publication bias in the papers used in this analysis was found using Egger's test (P = .55)(Supplementary Fig. S2)

Pre-Diagnostic Aspirin Use and All-Cause Mortality
A common effects model indicates that all-cause mortality is significantly higher in patients who took aspirin prediagnostically (HR 1.19, 95% CI, 1.11-1.27,P < .0001),whilst a random effects model shows a larger, but nonsignificant increase (HR 1.27, 95% CI, 0.95-1.72,P = .11)(Fig. 2b).Heterogeneity is again high (I 2 = 88%).Unlike breast-cancer-specificmortality, sensitivity analysis did not indicate that any one specific paper was responsible for this heterogeneity.Meta-regressions revealed multiple variables which were correlated with the magnitude of association.The proportion of patients treated with radiotherapy (n = 6, P < .0001)and the proportion of the cohort who were Black (n = 3, P = .0215)were both inversely correlated with an association of a greater magnitude.Meanwhile, the proportion of the cohort who were White (n = 4, P = .0054)and the proportion of the cohort who had never smoked (n =3, P = .0052)were both positively correlated with the magnitude of association (Supplementary Fig. S3).No evidence of publication bias was found (P = .66)(Supplementary Fig. S4).
Meta-regression revealed that later study year was associated with a reduced magnitude of association between postdiagnostic aspirin use and breast-cancer-specific death (n = 11, P = .0419)(Fig. 4).Another meta-regression showed that the percentage of patients in a study's cohort with node-positive cancer at diagnosis was inversely correlated with magnitude of association (n = 4, P = .0002)(Supplementary Fig. S7).
No evidence of publication bias (P = .44)was found in the papers used in this analysis (Supplementary Fig. S8).

Post-Diagnostic Aspirin Use and All-Cause Mortality
Fig. 5 shows the association between post-diagnostic aspirin use and all-cause mortality.There are small reductions in all-cause mortality in both common (HR 0.93, 95% CI, Abbreviations: pre, pre-diagnostic aspirin use; post, post-diagnostic aspirin use; BC death, breast-cancer-specific mortality; AC death, all-cause mortality; X, outcome assessed. The Oncologist, 2024, Vol. 29, No. 1 e7 0.87-1.00,P = .041)and random (HR 0.87, 95% CI, 0.71-1.07,P = .18)effects models, in addition to a large amount of heterogeneity (I 2 = 88%).Meta-regression revealed that the proportion of a paper's cohort who were black was positively correlated with an increased magnitude of association between post-diagnostic aspirin and all-cause death (n = 4, P = .0026)(Supplementary Fig. S9).There was no evidence of publication bias (P = .44)(Supplementary Fig. S10).

Post-Diagnostic Aspirin Use and Breast Cancer Recurrence
Fig. 6 shows the association between post-diagnostic aspirin use and breast cancer recurrence.Similarly to the results on the association of pre-diagnostic aspirin use with recurrence, there is no significant association between post-diagnostic aspirin use and breast cancer recurrence in both common (HR 0.91, 95% CI, 0.78-

Discussion
This meta-analysis looked at the association between prediagnostic or post-diagnostic aspirin use and breast cancer outcomes.It included 24 studies with a pooled cohort size of 149 860 patients with breast cancer.The study showed that pre-diagnostic aspirin use was not significantly associated with breast-cancer-specific mortality or recurrence, and appeared to be associated with slightly higher all-cause mortality.However, this latter finding is perhaps not unexpected given that a major reason for long-term aspirin use is in the secondary prevention of cardiovascular events, 41 and therefore patients taking aspirin could be expected to have a higher likelihood of poor health irrespective of breast cancer diagnosis.Meta-regression revealed a number of variables associated with the magnitude of association between pre-diagnostic  aspirin and all-cause mortality, including race, smoking status, and the proportion of patients in a cohort treated with radiotherapy.The proportion of a cohort who were Black being correlated with a reduced association between pre-diagnostic aspirin and all-cause mortality could potentially be explained by the fact that Black people have a higher overall mortality rate compared to White people in the USA, 42 where many studies included in this analysis were conducted, and by the fact that Black people have breast cancer features associated with worse prognoses more commonly than other races. 43dditionally, as smoking is linked to increased incidence of CVD, 44 it can be inferred that a greater proportion of smokers than non-smokers would be taking aspirin due to its use as a secondary preventative measure in patients with CVD.This, combined with the overall higher mortality rates in smokers, 45 could explain the reduced association between prediagnostic aspirin and higher all-cause death in cohorts with more never smokers.Finally, a greater proportion of patients in a cohort who were treated with radiotherapy being associated with reduced magnitude of association between aspirin and all-cause mortality could potentially be explained by an interaction between radiation-associated CVD 46 and patients taking aspirin pre-diagnostically as a result of existing poor cardiovascular health.
It should be noted that the small number of papers used in these meta-regressions limits the statistical power of these results.This is illustrated by the fact that the removal or inclusion of single papers was able to determine whether or not meta-regressions of other variables were statistically significant.For example, meta-regressions found that the proportion of a cohort's patients with stage IV cancer was significantly correlated with the magnitude of association between preor post-diagnostic aspirin and all-cause and breast cancer mortality, but in all 4 cases this finding was rendered nonsignificant when Fraser et al. 23 was removed from the analysis.In addition, the proportion of a cohort's patients with grade 2 cancer was significantly correlated with pre-diagnostic aspirin's magnitude of association with all-cause mortality only when Zhou et al. 34 was excluded from the meta-regression, but the proportion of a cohort's patients treated with chemotherapy was significantly associated with the same outcome only when Zhou et al.'s paper was included.These 2 papers had such a disproportionate effect on whether or not a variable was significantly correlated with magnitudes of association between aspirin use and breast cancer outcomes because they themselves found relatively strong associations between aspirin use and some outcomes, thus causing correspondingly large changes in the magnitude of association between said variables and breast cancer outcomes with their removal or inclusion from meta-regression calculations, especially when few other studies were involved in said calculations.
Our study also found that post-diagnostic aspirin use was not significantly associated with all-cause mortality or recurrence, but that post-diagnostic aspirin use appeared to be somewhat associated with breast cancer mortality.There are multiple potential mechanisms for this association.One hypothesis is that aspirin's inhibition of COX-2 leads to a beneficial effect, given that COX-2 overexpression has been linked to worse overall survival 2 in breast cancer.COX activity could lead to worse cancer outcomes through multiple pathways, such as promotion of the epithelial-to-mesenchymal transition 47 and induction of angiogenesis, 48 so inhibiting COX could improve breast cancer survival.Another potential mechanism by which aspirin could improve breast cancer survival is a reduction in cancer metastasis through inhibition of platelet function. 3,49,50he fact that post-diagnostic aspirin use was found to be significantly associated with lower breast-cancer-specific mortality, despite not being found to be significantly associated with lower disease recurrence, could be due to a smaller number of papers including data on the association between post-diagnostic aspirin use and breast cancer recurrence, with this smaller dataset making it less likely for any association to reach statistical significance.It could also reflect the fact that only 1 20 of the 4 18,20,21,23 papers with the strongest association between post-diagnostic aspirin use and breast-cancer-specific mortality included data on the association between postdiagnostic aspirin use and breast cancer recurrence.
The magnitude of the association between post-diagnostic aspirin and breast-cancer-specific mortality which this study finds is smaller than those seen in 3 previous metaanalyses, 6,7,9 although it is similar to that found in a more recent meta-analysis. 8As 2 previous meta-analyses were published in 2015, the difference between their findings and that of this paper matches the correlation between later study years and a reduced magnitude of association between post-diagnostic aspirin use and breast-cancer-specific mortality which we observed through meta-regression.Possible explanations for this correlation include the fact that papers published later tended to have a lower follow-up length, or the fact that the 3 papers 18,20,21 which used self-reported aspirin use data were the 3 earliest papers to look at the association between post-diagnostic aspirin and breast cancer mortality.Additionally, another relatively early paper with positive results 23 may have had these results influenced by its method of ascertaining aspirin use; as post-diagnostic aspirin use was assessed based on prescriptions up to the time of death, patients not having their prescriptions renewed when close to death 51 could have implied an erroneously strong association between aspirin use and breast cancer mortality.Our finding that post-diagnostic aspirin use in patients with ER-positive breast cancer was associated with a slightly stronger association with breast-cancer-specific mortality than in the study cohort as a whole is supported by a previous finding 52 that NSAID use was associated with reduced breast-cancer-specific mortality in patients with ER-positive cancer, and with increased breast-cancer-specific mortality in patients with ER-negative cancer.A possible explanation for this effect could be the finding that NSAIDs such as aspirin can reduce serum estrogen, 53 which could potentially reduce the growth of estrogen-responsive tumors.Finally, we also found that post-diagnostic aspirin use was more strongly associated with reduced breast-cancer-specific mortality in patients with lower-stage cancer at diagnosis.This suggests that aspirin may exert most of its effects in small, local cancers, perhaps through inhibiting tumor growth 54 and thus reducing the ability of the primary tumor to seed metastases. 55,56ur study's finding that pre-diagnostic aspirin use is not significantly associated with breast cancer outcomes is also seen in both prior meta-analyses 6,7 which looked specifically at pre-diagnostic or post-diagnostic aspirin use.These 2 meta-analyses also support this study's finding that postdiagnostic aspirin use is associated with a non-significant decrease in all-cause mortality.Only 2 prior meta-analyses 7,9 looked at the association between aspirin use and recurrence; one 7 finds that post-diagnostic aspirin use causes a larger decrease in recurrence than was observed by our study, whilst the second 9 finds similar results to this study.Conversely, a recent RCT, the ABC trial, 57 assessing whether post-diagnostic aspirin use can prevent recurrence in breast cancer found a small increase in recurrences in the treatment arm of the study compared to the control arm. 58verall, the results of our study appear similar to those from prior meta-analyses which looked at this subject, with the exception of a slightly smaller magnitude of the association being found between post-diagnostic aspirin use and breast-cancer-specific mortality.
Strengths of this meta-analysis include that it separates all results by whether aspirin was taken before or after diagnosis, that it includes subgroup analyses of variables that are not included in similar meta-analyses, and that it contains a number of meta-regressions that could help explain aspirin's associations with various breast cancer outcomes.
However, there are numerous limitations to this metaanalysis, not least the fact that the data that it includes is observational.First, the sample size for some meta-analyses is somewhat small, especially the analyses of aspirin's associations with recurrence.In addition to this, not all studies had data for all variables which were looked at via metaregression, so some meta-regressions were also limited by small sample sizes; this may explain why meta-regressions looking at stage or ER status did not show significant correlations, contrary to this study's meta-analyses which indicated that these variables affected aspirin's associations with breast cancer outcomes.In most analyses there tended to be a large amount of heterogeneity, which could be explained by a number of factors; one could be the large range of study follow-up lengths, whilst another could be the differences in aspirin use assessment between studies.Finally, it should be noted that all studies included in this meta-analysis were performed in wealthy nations in Europe or North America, with one study performed in Turkey 39 a potential exception.Caution must therefore be used in extrapolating these results to less wealthy nations where the standards of cancer treatment are often markedly different.The same point could be made about ethnicity, given that most countries in this study have majority White populations and different The Oncologist, 2024, Vol. 29, No. 1 e11 ethnicities known to have different cancer characteristics and outcomes. 43ur study finds that aspirin taken pre-diagnostically has little association with breast cancer outcomes.However, given that other meta-analyses 59 have found that pre-diagnostic aspirin use is associated with a reduced risk of a patient developing breast cancer, aspirin could still be a useful prophylactic option in populations at a high risk of developing breast cancer, such as those with a family history of the disease.It also finds that aspirin taken post-diagnostically may have a limited beneficial association with breast cancer mortality.Caution must be used when interpreting this result as the apparent significant association with breast-cancer-specific mortality is largely generated by 3 papers 18,20,21 which use self-reported aspirin usage data, and another paper 23 whose positive results may have at least partially been due to reverse causation.Given this, it may be prudent to wait for the results of the ADD-Aspirin trial 4 to draw firm conclusions about post-diagnostic aspirin use's association with breast cancer mortality.Aside from the fact that RCTs are generally viewed as a higher standard of evidence than observational studies, ADD-Aspirin has an arm consisting of patients taking 300mg of aspirin daily.Given that many of the studies in this meta-analysis involved patients taking low-dose aspirin, it could be that this arm of ADD-Aspirin generates a different result than this study found.
Should ADD-Aspirin, or other future RCTs, confirm this study's finding that post-diagnostic aspirin has some beneficial association with breast cancer mortality, then aspirin could be used as a long-term therapy in patients with breast cancer after primary treatment.Due to its low cost, aspirin could be especially useful in lower-income countries where more expensive breast cancer treatments are less readily available.Aside from its potential clinical uses, knowledge of the effects of aspirin on breast cancer outcomes could allow for its use to be identified as a confounding variable in other breast cancer studies.
Future studies could look for cohorts in whom aspirin is especially beneficial to allow its use to be more selectively indicated.They could look at variables such as ER status or stage at diagnosis, which were identified in this meta-analysis.Another, less well-studied variable that could be looked at is COX status.Given that aspirin is known to have effects on COX enzymes, it is surprising that there are few studies that analyze the effect of COX status on aspirin's association with breast cancer outcomes.Whilst it should be noted that one study 60 which looked at the relationship between COX status and aspirin's association with breast cancer outcomes found little difference between patients with COX positive or negative tumors, this study looked only at COX-2 status, so future studies could instead examine COX-1 status.Additionally, it could be useful for future studies analyzing the effects of aspirin on breast cancer outcomes to look at populations outside of Europe or North America, given that all of the studies in this meta-analysis were conducted in those regions.The ADD-Aspirin trial is already doing this, as it recruited a portion of its participants from India.

Figure 2 .
Figure 2. (a) Association between pre-diagnostic aspirin use and breast-cancer-specific mortality.(b) Association between pre-diagnostic aspirin use and all-cause mortality.(c) Association between pre-diagnostic aspirin use and recurrence.

Figure 6 .
Figure 6.Association between post-diagnostic aspirin use and recurrence.

Table 1 .
Key characteristics of included studies.

Table 2 .
Outcomes assessed in included studies.
Association between post-diagnostic aspirin use and all-cause mortality.