Aspirin in the Treatment of Cancer: Reductions in Metastatic Spread and in Mortality: A Systematic Review and Meta-Analyses of Published Studies

Background Low-dose aspirin has been shown to reduce the incidence of cancer, but its role in the treatment of cancer is uncertain. Objectives We conducted a systematic search of the scientific literature on aspirin taken by patients following a diagnosis of cancer, together with appropriate meta-analyses. Methods Searches were completed in Medline and Embase in December 2015 using a pre-defined search strategy. References and abstracts of all the selected papers were scanned and expert colleagues were contacted for additional studies. Two reviewers applied pre-determined eligibility criteria (cross-sectional, cohort and controlled studies, and aspirin taken after a diagnosis of cancer), assessed study quality and extracted data on cancer cause-specific deaths, overall mortality and incidence of metastases. Random effects meta-analyses and planned sub-group analyses were completed separately for observational and experimental studies. Heterogeneity and publication bias were assessed in sensitivity analyses and appropriate omissions made. Papers were examined for any reference to bleeding and authors of the papers were contacted and questioned. Results Five reports of randomised trials were identified, together with forty two observational studies: sixteen on colorectal cancer, ten on breast and ten on prostate cancer mortality. Pooling of eleven observational reports of the effect of aspirin on cause-specific mortality from colon cancer, after the omission of one report identified on the basis of sensitivity analyses, gave a hazard ratio (HR) of 0.76 (95% CI 0.66, 0.88) with reduced heterogeneity (P = 0.04). The cause specific mortality in five reports of patients with breast cancer showed significant heterogeneity (P<0.0005) but the omission of one outlying study reduced heterogeneity (P = 0.19) and led to an HR = 0.87 (95% CI 0.69, 1.09). Heterogeneity between nine studies of prostate cancer was significant, but again, the omission of one study led to acceptable homogeneity (P = 0.26) and an overall HR = 0.89 (95% CI 0.79–0.99). Six single studies of other cancers suggested reductions in cause specific mortality by aspirin, and in five the effect is statistically significant. There were no significant differences between the pooled HRs for the three main cancers and after the omission of three reports already identified in sensitivity analyses heterogeneity was removed and revealed an overall HR of 0.83 (95% CI 0.76–0.90). A mutation of PIK3CA was present in about 20% of patients, and appeared to explain most of the reduction in colon cancer mortality by aspirin. Data were not adequate to examine the importance of this or any other marker in the effect of aspirin in the other cancers. On bleeding attributable to aspirin two reports stated that there had been no side effect or bleeding attributable to aspirin. Authors on the other reports were written to and 21 replied stating that no data on bleeding were available. Conclusions and Implications The study highlights the need for randomised trials of aspirin treatment in a variety of cancers. While these are awaited there is an urgent need for evidence from observational studies of aspirin and the less common cancers, and for more evidence of the relevance of possible bio-markers of the aspirin effect on a wide variety of cancers. In the meantime it is urged that patients in whom a cancer is diagnosed should be given details of this research, together with its limitations, to enable each to make an informed decision as to whether or not to take low-dose aspirin. Systematic Review Protocol Number CRD42015014145

of eleven observational reports of the effect of aspirin on cause-specific mortality from colon cancer, after the omission of one report identified on the basis of sensitivity analyses, gave a hazard ratio (HR) of 0.76 (95% CI 0.66, 0.88) with reduced heterogeneity (P = 0.04). The cause specific mortality in five reports of patients with breast cancer showed significant heterogeneity (P<0.0005) but the omission of one outlying study reduced heterogeneity (P = 0. 19) and led to an HR = 0.87 (95% CI 0. 69, 1.09). Heterogeneity between nine studies of prostate cancer was significant, but again, the omission of one study led to acceptable homogeneity (P = 0.26) and an overall HR = 0.89 (95% CI 0. 79-0.99). Six single studies of other cancers suggested reductions in cause specific mortality by aspirin, and in five the effect is statistically significant. There were no significant differences between the pooled HRs for the three main cancers and after the omission of three reports already identified in sensitivity analyses heterogeneity was removed and revealed an overall HR of 0.83 (95% CI 0. 76-0.90). A mutation of PIK3CA was present in about 20% of patients, and appeared to explain most of the reduction in colon cancer mortality by aspirin. Data were not adequate to examine the importance of this or any other marker in the effect of aspirin in the other cancers. On bleeding attributable to aspirin two reports stated that there had been no side effect or bleeding attributable to aspirin. Authors on the other reports were written to and 21 replied stating that no data on bleeding were available. Chan et al (2009) [8], Langley (2011) [9,10] and others have pointed out that effects of aspirin on certain biological mechanisms justify an expectation of benefit from aspirin used as an adjunct treatment of patients with cancer. These effects include an interruption of tumour growth, a retardation of metastatic spread, an inhibition of angiogenesis, enhancements of both DNA mismatch repair and cellular apoptosis and an abrogation of invasiveness. Benefit from treatment with aspirin is therefore not unexpected.
Our aim in what follows is to provide a comprehensive systematic review and meta-analysis of the available evidence on the effects of aspirin used as an adjunct treatment of cancer in the reduction of mortality and metastatic spread.

Methods
The review protocol was registered in PROSPERO (registration number CRD42015014145). In reporting we have followed the PRISMA guidelines [11].
In December 2015 observational and interventional studies in Medline and Embase were searched using a pre-defined strategy with indexed descriptors and keywords including "aspirin", "acetylsalicylic acid", "cancer" "tumour", "neoplasm", "mortality", "death", "adverse effect", "bleed". The search was limited to human studies in peer-reviewed journals and conference abstracts. Reference lists of the included studies were also searched and recent conference proceedings scanned and topic experts contacted for additional studies.
Studies were selected for inclusion in meta-analyses if (a) the studied population comprised patients diagnosed with cancer; (b) aspirin was taken regularly after cancer diagnosis independently of whether it had been taken before diagnosis; (c) they were case-control studies, cohort studies or controlled trials; and (d) cause-specific mortality was available. All-cause cancer mortality, incidence of metastases and adverse effects were noted but were not criteria for selection.
Two reviewers independently excluded reports that did not meet inclusion criteria based on title and abstract. Full published reports were obtained for the remainder, and inclusion criteria were applied.
The origin of the patient group and other details in each report were examined, and if there appeared to be two reports based on the same patients, if the evidence required for the metaanalyses was not clear, or if important items were missing, the author(s) was contacted and asked for clarification. Authors of all the papers were also asked whether they had data on gastrointestinal or other bleeding, and if this had been a concern at any time. All these processes were conducted by one of the authors and were checked as appropriate by another author.
The methodological quality of the included studies was assessed and graded independently by two authors using the Newcastle-Ottawa Scale [12]. Differences in grading of reports on a nine point scale, were discussed and agreed.
Meta-analyses were conducted grouping the studies according to study design: intervention and observational studies. Subgroup analyses were conducted according to cancer types, key mutations and whether or not patients had taken aspirin only after diagnosis.
The summary statistics derived in the meta-analyses were either a hazard ratio or a risk ratio each with 95% confidence intervals. The analyses were carried out using the statistical package STATA. The inverse-variance method was used to weight the individual studies and provide the pooled estimate of effects. A 'random effects' model was used throughout to incorporate an estimate of between-study variation into the calculation of common effects. Funnel plots were created to highlight outlying studies and look at publication bias. Publication bias was assessed using Egger's test [13]. Sensitivity analyses were performed to assess the influence of individual studies on the combined hazard ratios. Heterogeneity was assessed using the Q statistic and investigated by repeating the meta-analyses excluding, first low scoring studies and then, if substantial heterogeneity was still present, outlying studies, identified by the sensitivity analyses were omitted.

Results
The literature search identified 373 reports and following omissions of duplicates and irrelevant reports, 42 were found to be relevant and gave sufficient data to be included in meta-analyses. We present a summary diagram in Fig 1 showing the selection process. Table 1 summarises a few basic features of the papers included in the meta-analyses together with those upon which further investigations are based. A final column gives an assessment of quality of the studies, judged according to the Newcastle-Ottawa Scale [12].
In the tables that follow we summarise the individual papers and report the results of metaanalyses and when available we give data for both cause-specific mortality and all-cause mortality.

Aspirin, specific and overall mortality
Our search identified four reports of randomised trials, together with a report of pooled trials [14] ( Table 2). The ad hoc trials [15][16][17][18] were small and the results did not achieve significance. The report by   [14] describes a 6.5 year follow-up of five early vascular trials, during which time 987 new cancers developed. In these, aspirin was associated with a reduction in cancer deaths (HR 0.71; 95% confidence limits (CI) 0.57, 0.90). The effect of aspirin in all five trials together is homogeneous (P = 0.30), but this result should be taken with caution as clinical heterogeneity such as differences between the design of the studies, the patient populations etc. may be too great to justify the pooling of results.
Six studies of other cancers [54][55][56][57][58][59] are included in Table 3. Benefit from aspirin is suggested in all six, but they are too diverse to justify meta-analysis.
A few of the reports give details of grade or stage of the cancer, but these were too few to enable any relevant analyses in relation to the effect of aspirin. Later however we quote a few comments on aspirin and 'advanced' cancer. Figs 2 and 3 respectively show the Forest plots of the cause specific mortality and the all-cause mortality (although only two reports [44,52] stated all-cause mortality in patients with prostate cancer).
It is possible to examine the pooling of the HRs for the three main cancers, and it seems not unreasonable to do this because the various pairs of HRs do not differ significantly (thus: for   colon and breast cancer P = 0.90; for colon and prostate cancer P = 0.06 and for breast and prostate cancers P = 0.18). An overall meta-analysis for cause-specific mortality from these three cancers is 0.78 (95% CI 0.66, 0.92). Naturally, this has to be accepted with great caution, Aspirin in the Treatment of Cancer particularly as there is significant heterogeneity (P<0.0005), even though the omission of the three papers already identified as 'outliers' by sensitivity analysis [20,38,40] reduces the heterogeneity (P = 0.03) and gives an overall HR of 0.83 (95% CI 0.76, 0.90). Egger's test for publication bias [13] is not significant (P = 0.30).

Aspirin and metastatic spread
An effect of aspirin on metastatic spread is clearly an evidence of treatment and the data in Table 4, although sparse, are therefore of considerable importance. Two studies of breast cancer [14,46], two of prostate [23,38] and one of both cancers together with colon [4], give evidence of a reduction in spread by aspirin. A combined estimate gives a relative risk for aspirin of 0.77 (95% CI 0.65, 0.92), though there is significant heterogeneity between the studies. The inclusion of NSAIDs other than aspirin posed difficulties but we assumed that aspirin was the major drug used, and evidence for this is given in one of the studies [36]. b The examination of heterogeneity by the omission of papers was based on sensitivity analyses. The Newcastle-Ottawa grade of Frazer et al [38] and Assayaq et al [44] were both 9/10 and correspondence with the authors revealed no likely reason the heterogeneity. The data for Bains et al (Grade 7/10) is taken from a poster presentation and details of adjustments for confounding appear to be limited. Attempts to correspond with the author failed. c Reimers [30] reported a risk ratio and could not be included in the meta-analysis.   Table 3.

Aspirin and mutations
A mutation in PIK3CA, a gene which produces a protein that increases Cox-2 and prostaglandin activity, has been shown to enhance the response of the tumour to aspirin. The prevalence of this mutation is stated in several of the present studies as around 15-20% [24,27,61]. Table 5 summarises the relevant data and confirms a marked reduction in mortality in tumours with the mutation (HR 0.45; 95% CI 0.28, 0.71), while it is uncertain if there is benefit from aspirin in patients without the mutation (HR 0.94; 95% CI 0.67, 1.32). This last statement is based on comparison between pairs of HRs using the normal approximation of the difference between log HRs. Evidence on these or other mutations appears not to be available in the present studies of breast or prostate cancers.

Aspirin taken before a cancer diagnosis
Does aspirin treatment affect cancers which have developed while aspirin has been taken? It may be that cancers which develop while aspirin is being taken are less responsive to the effect of aspirin. Table 6 summarises relevant data and shows that aspirin taken before the diagnosis of cancer is of little or no relevance to the treatment effect.

Advanced disease and high risk groups
Data in three reports of prostate cancer suggest that the effect of aspirin may be greater in advanced disease. Thus Daugherty et al [47] describe an effect of aspirin in 'advanced' prostate cancer (HR 0.37; 95% CI 0.15, 0.92) which is greater, than in localized disease (HR 0.86; 95% CI 0.47, 1.58) Similarly Jacobs, Newton et al [52] report an HR of 0.60 (95% CI 0.37, 0.97) in 'high-risk' patients, contrasted with the effect of aspirin in the total cohort (HR 0.98; 95% CI 0.74, 1.29) in the total series of patients. Neither of these differences are however significant, nor is a result reported by Jacobs, Chun et al [51] who selected 'high risk' patients and reported a reduction by aspirin HR 0.44; (95% CI 0.15, 1.28). Table 6. Aspirin also taken prior to diagnosis.

Author
Cancer Aspirin only after diagnosis, not before (95% CI) Aspirin after and before diagnosis (95% CI) The dose and consistency of aspirin taking Many of the reports state, or imply that aspirin at a dose appropriate for vascular protection had been used and only a very few reports comment further. Several studies report greater effects with 'high' dose aspirin [49,63] though no difference is significant. Several authors refer to the consistency of aspirin taking. Chan et al [8] give evidence consistent with a gradient (P < 0.04), the maximum benefit being with more than five aspirin tablets per week. Baastinnet et al [19] reported that the benefit of aspirin in all who took the drug as HR 0.77 (95% CI 0.63, 0.95), whereas 'frequent' use was associated with a possible slight increase in protection (HR 0.70; 95% CI 0.57. 0.88). Ng et al [29] reported HR 0.51 (95% CI 0.28, 0.95) for consistent use, compared with HR 0.68 (95% CI 0.42, 1.11) in the total cohort. Fuchs et al [25] state that compared with non-consistent use, 'consistent' users had a much greater reduction (HR 0.45; 95% CI 0.21, 0.97).
Several authors state that an effect of aspirin because apparent only after 3-5 years of therapy. Goh et al [26] state that they found evidence of benefit only after 5 years, Stock et al [53] who reported no benefit to prostate cancer overall (HR 1.03; 95% CI 0.79, 1.34) states that after five years of aspirin taking there was benefit (HR 0.54; 95% CI 0.26, 1.13) and an effect of aspirin taking in the study of lung cancer [55] became significant only after three years (HR = 0.84; CIs not stated).

Aspirin and bleeding
An excess of bleeding attributable to aspirin has been well studied in short-term vascular trials [64,65]. It is however appropriate to ask whether or not the risk of bleeding attributable to aspirin is similar in patients with cancer to that reported from the vascular trials. A few reports in the present series give a measure of reassurance on this. Din et al [23] who examined NSAID use state that there were no major bleeding complications. Liu et al [18] state that no side effects caused by aspirin were noted in any patient in the study. Curigliano et al [66] examined short-term aspirin taking by patients with breast cancer and stated that no major bleeding complication occurred.
The corresponding authors of the other reports in this series were written to. Replies received from twenty-one authors stated that no data on bleeding had been recorded.

Discussion
The evidence we present from a systematic overview of the literature gives support to the use of aspirin as an additional treatment of cancer. The evidence is limited, and while it is encouraging in the case of bowel cancer, there is insufficient evidence to dismiss a role for aspirin as an adjunct treatment of cancers other than colorectal. In fact, its use can be justified on the basis of its likely benefit on outcomes other than death, including its probable reduction in metastatic spread and its reduction in vascular disease events, including venous thromboembolism.
Differences between individual studies leading to significant heterogeneity is to be expected in any collection of observational studies such as those we present, and it does limit confidence in the results. However, if, for each of the three cancers, an out-lying study identified by detailed sensitivity analyses is omitted, heterogeneity is reduced to an acceptable level and for each cancer there is evidence suggestive of reductions in mortality and in metastatic spread.
In colon cancer there is evidence of a reduction in colorectal deaths of about 25%, and perhaps about 20% in All-cause mortality. If one report [13] is omitted the evidence of benefit in breast cancer is of a possible 13% reduction in cause-specific deaths, and for prostate cancer a possible reduction of perhaps about 11%.
With the present level of evidence, the pooling of data for the three cancers would seem to be not unreasonable and following omissions of three outliers, unacceptable heterogeneity is resolved. A meta-analysis then suggests a possible overall reduction by aspirin of about 15% (pooled HR 0.83; 95% CI 0.76-0.90). The evidence of a reduction in metastatic spread (RR 0.77; 95% CI 0.65-0.92) gives further encouragement to the use of therapeutic aspirin in cancer while awaiting evidence from ad hoc randomised trials.
It would be unreasonable to attempt to draw firm conclusions from the single studies on lung cancer [31], on oesophageal cancer [45], and on lymphatic leukaemia [22]. However the suggestive benefit in these studies, together with that reported for a mix of colon and women's cancer [21] indicates an urgent need for more observational studies in the less common cancers, some of which may never be subjected to evaluation in randomised trials. It would also be helpful if in the reporting of new studies information on the stage and grade etc. of the cancers could be indicated.
The evidence we present on the biomarker PIK3CA has been confirmed in an overviews by other authors [67], and similar data on other markers have been shown by other authors [68]. Thus Sun et al [31] who examined CTNNB1, a gene associated with cell adhesion and of relevant to familial polyposis, reported a marked enhancement in the effect of aspirin (HR 0,53; 95% CI 0.30, 0.95) compared with the effect in patients with the wild gene (HR 1.06; 95% CI 0.62, 1.83). A different approach to this issue was adopted by Chan et al [8] who used an overexpression of COX-2 in the primary tumour as an indication of a relevant mutation, and showed that overexpression was associated with reductions in colon cancer mortality (HR 0.39; 95% CI 0.20, 0.76) compared with the effect in other patients (HR 1.22; 95% CI 0. 36, 4.18).
All this suggests that the reduction by aspirin may be restricted to patients whose tumours show mutation in PIK3CA, HLA class I antigen, or show COX-2 over-expression. In colorectal cancer these subgroups represent approximately 17%, 54% and 50% of all patients, and our data suggest a reduction of about 50% in colorectal mortality, though another overview [67] suggested a reduction of only about 30%, while neither overview showed any reduction in those without the mutation. The scarcity of evidence on mutation in cancers other than colon is most unfortunate [69].
And yet any selection of patients for treatment with aspirin on the basis of a mutation or any other marker of cancer risk would be totally unwarranted on present evidence. Metastases are a major source of pain and other undesirable effects in solid cancers [70,71] and perhaps 90% of cancer deaths are at least in part due to metastases [72] and the withholding of aspirin would deny these possible benefits. Furthermore, the risk factors for vascular disease overlap with those for cancer and the withholding of aspirin would also deny patients the vascular benefits of aspirin, including the possible reduction of the excess risk of venous thromboembolism during chemotherapy. In fact, the marked increase in the risk of venous thrombosis in patients with cancer [66,73], has been shown to be reduced by low-dose aspirin [74], and it has therefore been recommended that prophylactic anticoagulants should be considered in all patients with cancer [75].
A major uncertainty in what we report arises from the possible omission of relevant reports, together with publication bias, and the test we performed suggests that this last may have occurred (Egger's test [13] P = 0.037). Furthermore, underlying all observational studies is the issue of residual confounding, and while this cannot be dismissed, it seems unlikely to have operated to any important degree as all the studies reviewed included multivariate adjustments.
On the other hand certain time biases could be present in some of the studies, and especially in retrospective case-control studies [76]. Patients not taking aspirin at the time of diagnosis can be defined, used as 'controls' and followed thereafter. Patients who start taking aspirin after receiving a diagnosis of cancer cannot be identified as 'cases' until they start taking the drug. It is possible that these will be identified later than the 'control' patients, and they will therefore be observed and deaths identified during a shorter time that that during which the patients not taking aspirin are observed. This has been called an 'immortal' time bias and Assayag & Azolay [44] include a detailed discussion of it.
All the reports in the present series were examined and while immortal time bias cannot be dismissed with certainty, an important effect upon the overall estimates of the effect of aspirin seems most unlikely. In fact, there is little difference in the overall mortality of the patients who had taken aspirin before diagnosis, and (presumably) continued to take it after diagnosis (HR 0.92), and the patients who had not taken aspirin before diagnosis (HR 0.90), in whom there could have been a time lag and thus, an 'immortal time bias' (see Table 6).
While a serious limitation in the present evidence is that little comes from randomised trials, yet evidence from further observational studies is urgently needed to evaluate more fully how patients likely to benefit from aspirin can be identified. In particular evidence on PIK3CA, other mutations and other possible markers should be collected as a matter of urgency in cancers of breast, prostate and other organs. The results of such studies should be made available for the encouragement and guidance of colleagues setting up randomised trials, and, in fact, the further question arises whether or not these mutations are of relevance to aspirin used in prophylaxis.
The possible benefits of aspirin must of course be evaluated against it side effects. Shortly after aspirin taking commences the risk of a gastrointestinal (GI) bleed is high but the risk falls rapidly thereafter [77,78], and in short-term trials the additional risk of a bleed from low-dose aspirin amounts to perhaps one or perhaps two patients in every 1,000 on low-dose aspirin [64,65]. After about three years of aspirin taking however, there appears to be no evidence of any excess GI bleeds attributable to the drug [78]. Moreover, the incidence of GI bleeding is highly sensitive to the presence of gastric pathology [78,79], and careful enquiries should therefore be made about current or past gastric symptoms, and about a high alcohol consumption [80]. The use of a gastroprotective drug together with the aspirin should be carefully considered if pathology is suspected.
The most serious bleeds are those that lead to death, and despite frequent references to fatal bleeds attributed to aspirin, there appears to be no valid evidence that deaths from GI bleeds are increased by low-dose aspirin [81]. In a recent study of patients admitted to hospital with 'gross' GI bleeding [82] the hospital stay of patients who had been taking aspirin was significantly shorter than that of patients who had not been on aspirin, and no patient whose bleed had been attributed to aspirin experienced an uncontrolled haemorrhage or died due to excessive bleeding.
Cerebral bleeds attributable to aspirin are rare, about one or two per 10,000 patient-years. Hypertension is the major factor in such bleeds [2] and in a randomised trial of aspirin based upon patients with hypertensive disease all of whom were adequately treated with anti-hypertensive drugs, there was the same number of cerebral bleeds in ten thousand patients on aspirin (19 patients) as in ten thousand on placebo (20 patients) [83]. A reduction in the risk of a cerebral bleed is therefore likely if the blood pressure of every person starting aspirin is checked, and adequately treated if raised [84].

Other overviews
An early overview of studies of aspirin and cancer was based on three small randomised trials and two observational trials, and this led to the conclusion: 'aspirin may have a role in the adjuvant setting. . . and should not be overlooked [9]. A further overview of three studies of colorectal, one of breast, two of prostate and one of oesophageal cancers judged that aspirin decreases the development and spread of metastases [10].
Li et al [85] reported a systematic review and meta-analysis of seven studies of colorectal cancer. Aspirin use after a diagnosis of colon cancer was reported to reduce overall all-cause mortality (HR 0.84; 95% CI 0.75, 0.94), but if aspirin had been taken before cancer benefit was uncertain (HR 0.77; 95% CI 0.52, 1.14). Ye et al [86] conducted a meta-analysis of seven of the studies of colon cancer in the present review, and reported effects of aspirin on both colon cancer mortality and overall mortality very close to what we report.
Huang et al [87] identified 16 studies in which a NSAID or aspirin had been used in patients with breast cancer. On the basis of meta-analyses they reported that aspirin taken after diagnosis, but not before, was associated with improved breast cancer survival (HR 0.69; 95% CI 0.50, 0.96). Zong et al [88] also reported a systematic review and meta-analysis of breast cancer patients in eight cohort studies and two nested case-control studies, and judged that post-diagnostic aspirin was associated with a significant reduction in the relative risk of death from breast cancer (RR 0.84; 95% CI 0.63, 1.12). Liu et al [89] conducted a systematic review and identified 39 studies of NSAIDs, including aspirin. A meta-analysis of prostate specific mortality in seven studies of aspirin gave an HR = 0.86 (95% CI 0.78, 0.96) and the effect of aspirin in these studies were judged to be more consistent than those for other NSAIDs.

Conclusions
It appears likely that low-dose aspirin has a beneficial role as an adjunct treatment of cancer. Reductions in mortality are shown in colon cancer, probably in prostate cancer and possibly in breast and individual studies of several other cancers also suggest benefit. Aspirin benefit in colorectal cancers, and possible other cancers, may be restricted to patients with tumours expressing certain genetic mutations. However, other benefits of low-dose aspirin, including reductions in metastatic spread and in vascular events, including venous thromboembolism appear to be independent of these biomarkers, and so information on aspirin should be given to patients whatever the state of the possible biomarkers.
The heterogeneity within the currently available studies-both between different cancers, and within the different studies of each cancer, together with evidence suggesting some publication bias, are such that further evidence from a number of adequately powered randomised, placebo controlled trials is urgently required, including trials of less common cancers. Evidence on the possible role of aspirin in uncommon cancers, and the possible enhancement of its effect if mutation and other markers of increased sensitivity to the actions of aspirin, are also urgently needed. Much of this evidence could some from further observational studies.
Nevertheless, despite the need for randomised trials, we believe the evidence of benefit from aspirin is sufficiently persuasive that physicians should engage with patients in a presentation and discussion of aspirin as an additional treatment. Furthermore, we hold that patients should be given this evidence within the context of a healthy lifestyle [90], they should be allowed to make their own decision about aspirin therapy, and should then be supported in whatever decision they make [91].
Supporting Information S1

Author Contributions
Conceived and designed the experiments: PCE GM SD. Analyzed the data: JEP JG AW DM MK SD. Wrote the paper: PCE JEP JG GM MK SD. Was involved throughout the study, in analysis, conducting literature searches, identifying studies for inclusion, and in writing the manuscript: AW.