A large body of evidence spanning preclinical, observational, and clinical data supports the potential use of nonsteroidal anti-inflammatory agents (NSAIDs) for cancer prevention, and possibly as adjuncts to cancer treatment. NSAIDs hold great appeal in this regard because they are relatively safe and inexpensive, and aspirin offers the potential to prevent, treat, or palliate several common, chronic, age-related diseases (eg, prevention of cardiovascular events, cancer risk reduction, treatment of inflammatory conditions, pain management). To date, data supporting NSAID use for cancer risk reduction have been most compelling and consistent for colorectal cancer (CRC), where celecoxib and aspirin have shown efficacy in reducing the incidence of precancerous polyps and in reducing polyp/CRC incidence and mortality, respectively.^[1–3]

Like CRC, breast cancer is a common cancer, with documented overexpression of COX-2 and for which there is a need for more effective and affordable preventive and treatment strategies. Observational data are mixed but generally suggest that aspirin may improve breast cancer survival.^[4] Not surprisingly, data from large randomized controlled trials with breast cancer–specific outcomes are currently lacking for both aspirin and celecoxib. In this issue of the Journal, Strasser-Weippl et al. report on findings related to the use of celecoxib and aspirin in the National Cancer Institute of Canada Clinical Trials Group MA.27 trial.^[5] The authors conclude that neither celecoxib nor aspirin demonstrated statistically significant effects on event-free survival or distant disease–free survival; however, low-dose aspirin administration was associated with lower overall survival. Unfortunately, there are a number of limitations to this study that make it difficult to draw definitive conclusions regarding celecoxib’s efficacy or the effect of aspirin on breast cancer or overall mortality.

First, regarding the effect of celecoxib on event-free survival (primary end point), conclusions are based on a much smaller sample size than initially planned, given that celecoxib was discontinued early due to the National Cancer Institute’s warning about adverse cardiac events. Consequently, while 811 individuals were randomly assigned to celecoxib, only 735 were actually exposed to it, with a median duration of exposure of just 4.6 months. Additionally, compliance is unclear but potentially concerning, considering the 32% discontinuation rate reported for the use of the aromatase inhibitors in this trial.^[6] Given the weakened statistical power resulting from the curtailed sample size, the limited median duration of celecoxib exposure, and the limited mean follow-up time, results with regard to the effect of celecoxib on breast cancer event–free survival, distant disease–free survival, and overall survival can only be considered inconclusive. While trials examining the effect of celecoxib on colorectal precancers have seen effects as early as one year,^[2–3]it is likely that a much longer duration would be needed to see meaningful effects on a clinical outcome like invasive cancer. Unfortunately, no intermediate biomarker data were available from the trial that could have provided some insight into the biologic effects of the agents that may have helped clarify the underpowered results.

With respect to the authors’ conclusions regarding the effect of aspirin on clinical end points, several caveats limit the ability to interpret these results. First, the study of aspirin use by participants in this clinical trial is, in effect, an observational study, with aspirin use being a stratification factor rather than a randomized variable; therefore, outcome analyses of aspirin users vs nonusers are highly likely to be biased. The authors report no effect—or more accurately, a lack of association—of aspirin on either event-free survival or distant disease–free survival in multivariable analyses. But given that just 21.5% of participants reported low-dose aspirin use, one cannot exclude the possibility that the trial was simply underpowered to detect an effect of aspirin on these outcomes over a relatively short follow-up period. However, Strasser-Weippl et al. report reduced overall survival in those taking aspirin. They suggest that this is likely not an effect of aspirin itself, but due to the increased baseline risk of cardiovascular mortality among those taking aspirin. While this may be the case, the authors do not provide any data regarding causes of death by aspirin status. They report that there were 56 deaths due to any cause in those using aspirin and 110 in those not taking aspirin, but what proportion of each of these was due to breast cancer? What proportion was due to cardiovascular disease? A prior publication from this trial suggests an increased risk of non–breast cancer death among participants but does not provide data by aspirin status.^[7] Furthermore, the authors suggest that some previous observational studies have not properly acknowledged that there are likely differences in those who take aspirin before a cancer diagnosis (increased cardiovascular disease [CVD] risk) and in those whose only use occurs after a cancer diagnosis (cancer protective effect). However, the data surrounding aspirin for the prevention of colorectal and total cancer incidence and mortality, as well as all-cause mortality, have evolved considerably since the start of this trial, and one wonders if the distinction between pre- and postdiagnostic use today is as relevant as it once might have been. Aspirin use is already common in the population, particularly among older adults, and is likely to become more so with the recent US Preventive Services Task Force (USPSTF) recommendation for the use of aspirin to prevent colorectal cancer in addition to CVD in those at increased cardiovascular risk.^[8] This suggests that the influence of aspirin on breast cancer recurrence in isolation may be less relevant today than it previously was. Today, the more salient questions regarding aspirin use are related not only to cancer risk reduction, but to a wider range of cardiovascular- and cancer-related outcomes, dose of aspirin, safety, and to issues of patients’ quality of life more broadly, rather than the effects of aspirin on any single cancer in isolation.

The results of the UK REACT trial (NCT02429427), the ABC trial (NCT02927249), and the Add-Aspirin trial (NCT02804815) will be critical to determining the true effect of celecoxib and aspirin on breast cancer survival and mortality. Importantly, these studies are randomized clinical trials (as opposed to observational studies) that include overall survival and safety outcomes.

The potential for NSAIDs to reduce cancer incidence and mortality is tremendous and has been suggested for decades. The field has come closer to realizing this potential with the incorporation of an indication for colorectal cancer prevention in the updated USPSTF recommendation for aspirin use in those at an increased CVD risk.^[8] Unfortunately, the impact of the celecoxib and aspirin findings from the Strasser-Weippl et al. study are limited. Additional data from ongoing randomized controlled trials of aspirin in other cancers and other common diseases and in various populations^[9–11] will be needed to determine when, where, and how to apply NSAIDs to achieve their maximum benefit in the prevention of cancer and other common chronic diseases.

Funding
Supported in part by the Boone Pickens Distinguished Chair for Early Prevention of Cancer to E. Hawk and by the MD Anderson Cancer Center Support Grant 5P30CA016672-37 from the National Institutes of Health.

J Natl Cancer Inst. 2018;110(9):927-928. © 2018 Oxford University Press