Thursday, 23 December 2010

Crabsallover comments on Moayyedi & Jankowski review of Peter Rothwell Aspirin effect on Cancer.

reposted from:

Unedited full version of Moayyedi & Jankowski here.

crabsallover comments & calculations in blue, highlights in red, keypoints in bold red.

paid £23 - stored on crabsallover PC: Moayyedi and Jankowski 341bmj-rothwell-aspirin-cancer.pdf

BMJ 2010; 341:c7326 doi: 10.1136/bmj.c7326 (Published 22 December 2010)
Cite this as: BMJ 2010; 341:c7326
  • Editorial

Does long term aspirin prevent cancer?

  1. Paul Moayyedi, professor of medicine1
  2. Janusz A Jankowski, James Black fellow2
+Author Affiliations
  1. 11200 Main Street West, Hamilton, ON, Canada
  2. 2Division of Clinical Pharmacology, Department of Oncology, University of Oxford, Oxford OX3 7DQ, UK
Potentially, but trials specifically designed to answer this question are needed
The cardioprotective effects of aspirin are well established. A meta-analysis of individual subject data from primary prevention randomised controlled trials (RCTs) suggested that aspirin can reduce the relative risk of non-fatal myocardial infarction by about 20%.1 Overall, however, the risks of treatment (severe gastrointestinal and extracranial bleeding) were roughly the same as the benefits, so routine use of aspirin as a primary preventive strategy was not recommended. 
For Primary Prevention note the +0.03% (116 cases) increase in major extracrannial bleeds & +0.01% haemorragic stroke (27 cases) with aspirin and the -0.05% (160 cases) & 21% (596 v 756 absolute) reduction in non-fatal myocardial infarction (MI)

Aspirin seemed to increase the incidence of haemorrhagic stroke both in the primary and in the secondary prevention trials. Conversely, aspirin seemed to reduce the incidence of ischaemic stroke in both types of trial.  
Ischaemic strokes are those that are caused by interruption of the blood supply, while haemorrhagic strokes are the ones which result from rupture of a blood vessel or an abnormal vascular structure.

The meta-analysis did not evaluate any potential reduction of mortality from cancer, as has been suggested by observational data.2 Observational data are difficult to interpret, however, because associations may not be causal and may be the result of confounding or bias.3 Rothwell and colleagues have therefore conducted another meta-analysis of individual subject data from RCTs of aspirin versus no aspirin for prevention of vascular disease, but this time they evaluated mortality from cancer as the main outcome.4 They found a 21% (95% confidence interval 8% to 32%) reduction in the odds of death from cancer in people who took aspirin for almost six years, and the effect was strongest for gastrointestinal cancers. The authors suggest that this may now tip the balance in favour of using aspirin as a primary prevention strategy, both for ischaemic heart disease and cancer.
Cordelia Molloy/Science Photo Library
The data initially look compelling, and primary care providers could be forgiven for now advising patients above a given age to take aspirin. Recommending widespread use of a drug in an otherwise healthy population, however, needs to be approached cautiously because we need to be certain that the benefits will outweigh the risks.1 A 20% reduction in death from cancer seems impressive, but looking at the data in terms of absolute numbers gives a different story. These trials followed up 25 750 participants for more than 150 000 patient years, and 20 fewer cancers occurred in those taking aspirin compared with those not taking aspirin. 
Looked at another way, after taking aspirin for almost six years there was a 0.5% absolute reduction in death from cancer, with a number needed to treat of 200 (111 to 1075). 

David Spiegelhalter at Understanding Uncertainty illustrated the 'number needed to treat' for statins was 50. In his 'spinning the risk' select the drop down for 'statins 10 yr' (not bacon sanwiches!). Now select 'No. needed to treat' box: 'In order to save one person from experiencing a heart attack or stroke in 10 years, we would need to treat 50 people like you with Statins.'

Whether the figure is 200 for aspirin or 50 for statins the net benefits are potentially life-saving.  

Compared with the 50% increase in risk of serious gastrointestinal and extracranial bleeding in participants taking aspirin (overall more than 100 excess serious bleeds occurred in the aspirin group in primary prevention trials),1 it is not clear that the benefits outweigh the risks, even when factoring in the cardioprotective effects of aspirin.
In a podcast Nigel Hawkes  spoke about absolute and relative risks (25'), benefits expressed as a % ... but risks or disbenefits are often expressed as numbers eg 1 in 2000 - which sound small. This is called 'mis-matched framing'.

Here Moayyedi & Jankowski have used a version of 'mis-matched framing viz. comparing absolute numbers of deaths saved by aspirin (20) to the absolute number of serious bleeding (100). They compare deaths with serious but non fatal illnesses and then say 'it is not clear that the benefits outweigh the risks, even when factoring in the cardioprotective effects of aspirin.'

Peter Rothwell has said that only 5% of serious bleeds results in death so of the 100 excess serious bleeds in the Antithrombotic Trialist’ (ATT) Collaboration trial they cite (not the Peter Rothwell trial), we'd expect only extra 5 deaths. 

2.18'- 2.19': Fergus Walsh: But there are side affects and they can be serious?
2.20'-2.55': Peter Rothwell: They can. Their is a risk bleeding, of major bleeding of 1 per 1000 years for each patient OR 1 per 1000 patient years OR 1 bleed per 1000 patients per year. Only a small proportion about 5% of those bleeds are fatal, so it's a small risk of death. When you put together all the risks, the risk of bleeding, the risk of heart attack and stroke and the effect on cancer, from the trials we looked at, taking aspirin reduces the risk  of premature death by about 10%.

The most rigorous approach to assessing this is to evaluate all cause mortality. To their credit, Rothwell and colleagues analysed this in some detail and report a statistically significant reduction in all cause mortality.4Crucially, the effect is only marginally significant (relative risk of death in the aspirin arm 0.93, 0.87 to 1.00; P=0.045) (figure). 
P=0.045 ie 1 in 22 risk (1/0.045) that result was by chance. The relative risk of death in the aspirin arm of 0.93 =7% reduction in death risk.
This is not particularly robust, especially as six of the eight trials were in subjects at high risk of cardiovascular disease, where the benefits of aspirin should be greatest. Furthermore, the authors included one RCT where the data had been destroyed.5 

Rothwell et al (pg 4) states 'Individual patient data were available from seven trials, but all records of one trial had been destroyed (Juul-Moller S, University Hospital, Malmo, Sweden, personal communication)..ref: 22'

It could be argued that because they had intended to conduct a meta-analysis of individual participant data this trial should have been excluded. This trial (SAPAT) had the greatest effect on all cause mortality (figure),5 and when it is excluded the results are not significant (0.94, 087 to 1.01; P=0.11). 
P=0.11 ie 1 in 9 (1/0.11) risk that result was by chance
This is corroborated by another meta-analysis of individual participants (95 000 subjects with 660 000 person years of follow-up), which found that aspirin had no significant effect on all cause mortality (0.95, 0.88 to 1.02; P=0.1) when used as primary prevention.1
All Cause Mortality:
1432/14035 (aspirin) 10.2% v 1275 /11535 11.05% (control)

As Peter Rothwell says 'The reduction in cancer deaths on aspirin during the trials resulted in lowered in-trial all-cause mortality (10·2% vs 11·1%, OR 0·92, 0·85—1·00, p=0·047, webappendix p 4)'.

Moayyedi & Jankowski's modified table:-
Forest plot of all cause mortality in randomised controlled trials of aspirin versus no aspirin4

Rothwell: DOI:10.1016/S0140-6736(10)62110-1
The Rothwell paper shows the 20 excess cancer deaths in control groups (327 out of 14035 with aspirin v 347 out of 11535 with control). The Moayyedi & Jankowski table shows 'all-cause deaths or 'events' rather than just cancer deaths eg 1432/14035 with aspirin and 1275/11535 with control. 
If 15 million UK persons (everyone between 45 and 70 years) took aspirin then according to the Peter Rothwell paper I calculate that 11,650  (15,000,000 x 20/25750) people would not die. cf. crabsallover earlier calculation of 3,900 lives saved. (crabsallover 15th December 2010 calculations)

This last point emphasises the fact that Rothwell and colleagues chose a subset of all the primary and secondary prevention aspirin RCTs that have been conducted. Twenty one RCTs have evaluated about 112 000 subjects,1so Rothwell and colleagues evaluated only around 20% of all subjects. This is because they excluded several trials that did not meet their eligibility criteria (for example, insufficient length of follow-up or subjects given aspirin on alternate days). 

Rothwell et al states (pg 2) that "Selection criteria was for randomised trials of aspirin versus control that had a mean or median scheduled trial treatment period of at least 4 years and a range extending beyond 5 years."

However, systematic reviews can give different conclusions depending how they are conducted, and eligibility criteria can be somewhat arbitrary.6 For example, Rothwell and colleagues report that 75 mg aspirin a day is sufficient to reduce death from cancer. If this is the case, then perhaps 325 mg every other day would also be effective, as used in the US Physicians Health Study,7 which had 22 071 participants (almost as many as Rothwell and colleagues’ entire meta-analysis). It would be interesting to see how the inclusion of this trial would affect the results.


Peter Rothwell hypothesis is that a daily aspirin is required, not aspirin taken every other day, whatever its dose. The RCTs were chosen based on this hypothesis.
The data are not robust enough to recommend aspirin for the general population and we still do not know the optimum dose and length of treatment.8 Key data could come from the AspECT trial,9 10 which is evaluating the effect of aspirin in preventing all causes of mortality and oesophageal adenocarcinoma. The interim analysis will be reported in 2012.
In addition, Canada and the United Kingdom have low numbers of gastroenterologists per head of population compared with other countries in the developed world.(11) If most of the population started taking aspirin our gastrointestinal services might be overwhelmed. 

Until further evidence from RCTs is accrued, aspirin should not be recommended for everyone.

Crabsallover is a 54 year old male in apparently good health. I conclude, after this Jankowski & Moayyedi article, that the benefits of reduced risk of cancer, heart attack and stroke, by taking daily 75mg aspirin, for me, do outweigh the increased risks of death caused by bleeding and brain stoke. I started taking daily 75mg aspirin on 8th December 2010 after the Peter Rothwell paper was published but I will review this practise when further evidence is available about the risks and benefits of aspirin or when I have bleeding symptoms etc.

Cite this as: BMJ 2010;340:c7326


  • JJ had support for this work from Cancer Research UK, PM had no support; JJ is a consultant for Astrazeneca Oncology and PM and JJ have received Astrazeneca educational grants; JJ is chief investigator and PM is deputy chief investigator of the AspECT trial.
  • Provenance and peer review: Commissioned; not externally peer reviewed.


No comments: