Monthly Archives: June 2011

Week 4 – Summary: Medical schools and Performance in MRCP

When I was introducing this paper I chose to highlight that one of the reviewer’s thoughts that it was a poor quality study. I don’t know  if that influenced the discussion or even non-participation in last week’s #twitjc, but there were several tweets expressing disappointment with the paper during the discussion. At first glance this appeared to be an accessible paper on medical education which would provoke a lot of discussion. But when you look more closely it contains complex analyses; the results of which many of those reading the paper did not manage to get to grips with. The poor presentation of some of the results in the additional file did not help. And the authors reach conclusions which are hard to justify.

Overall the main finding of the research was that medical schools seemed to have little impact on how students performed in post-graduate examination. The better the intake of students were at passing exams at 18, then the more likely they were to pass exams a decade later. This prompted me to ask if that suggested that rather than a national exit exam we needed a national entrance exam. @twsy suggested that if we wanted to look at the ‘value added’ by the medical school then we would need a national entrance and exit exam.

Some medical schools have graduates who take longer to pass professional exams. Is this an issue that should concern medical schools? And if it is what should we do about it? The correlation demonstrated in this paper suggests that if we wanted to have uniform outcomes for graduates of all medical schools then we should have uniform intake. It is unlikely that it would be socially acceptable to make students complete a national entrance exam and then allocate them to medical students across the UK to ensure an equal mix of academic performance. So we are left with the current situation.

We asked if performance of graduates in post-grad exams was a good indicator of the performance of a medical school. We didn’t think that it was, but we weren’t sure how performance of a medical school should be assessed, or if it should be at all. As an aside there was some discussion about what would make a good doctor at the individual level. Was it ‘head knowledge’ or good communication? It was pointed out that UKFPO was now trialling an assessment of situational judgement as a way of allocating doctors to further training. This is certainly something I would like to learn more about.

Access to the reviewers’ comments was generally lauded. We would like to see this more, as it can help understanding of the paper. In my own opinion it would be interesting to have seen some editorial comment on how two such different reviews were rationalised so that the outcome was that the paper was published.

I know that some people missed out on participating in this discussion, so I hope that you will take the opportunity to leave a comment here.

Should we have discussed this paper? Yes, we should. Many people will have heard of it before, and now they hopefully have a better understanding of it’s findings and linitations. #win!

Posted in Summaries, Week 4 | 4 Comments

Week 4 – Introduction and Discussion points: Medical schools and Performance in MRCP

Thank you Fi for asking me to write an introduction and suggest some discussion points for tonight’s Twitter journal club. Having a special interest in medical education I was very happy to see this paper by McManus et al. suggested for Week 4 of the journal club.

Discussion point 1: What factors do you think might explain variation in performance in MRCP between medical schools?

What did authors do? They looked at outcomes in MRCP (Member of the Royal College of Physicians) examination for entrants from all medical schools between 2003 and 2005. They found that in the Part 1 and 2 exams, Cambridge, Oxford, and Newcastle graduates did significantly better than average , and the performance of Liverpool, Aberdeen, Dundee and Belfast students was significantly worse. In the PACES section (a clinical examination based on a modified OSCE) Oxford students performed significantly and Liverpool, Dundee and London students significantly worse.
This first part of the analysis is quite easy to understand but the authors then go on to construct a multi-level model to see if they can explain variation between the medical schools.
Since it is known that ethnicity and gender are correlated with MRCP performance, and they had this as individual level data, they adjusted for this.

Discussion point 2: Is it surprising that the average offer to those applying to a medical school may predict performance of graduates in MRCP?
Two complex analyses were performed in this study: a multilevel model, and a structural equation model. Unfortunately the results of the multilevel model are not produced in an easy to understand format, although there is a figure in an additional file which is downloadable.
The authors looked for correlations between medical school performances in MRCP and a plethora of other factors. This information was pulled from other sources such as the Guardian tables, a survey of the cohort of medical students who started university in 1990/1, and the offers which each medical school made to students in the mid-1990s.
They found correlations between:
- Offers made to students (A level or Scottish higher grades)- the higher the offer the better the performance
- The proportion of final year medical students reporting being interested in a career as a physician, and reporting interesting medical teaching, and better performance in MRCP.
- The higher the percentage of graduates taking MRCP, the better the performance.
However when these factors were analysed together, it was only admission grades that seemed significant.
They also looked at correlations with data from the Guardian tables. In a multiple regression, again only admission criteria were found to be significant.
In the multilevel model, the entrance qualifications of graduates were found to explain 62% of variance, which in this type of study is a large amount. 38% of variance remained unexplained but a commenter has suggested that the contribution of entrance qualifications may be under-estimated because of the ‘ceiling effect’- many entrants may have been offered the highest grades of 3 As.

Discussion point 3: Are the authors correct to conclude that this analysis suggests that a national exit exam should be introduced?
What do the results of this study mean? To place the study in context it is perhaps useful to start with the last words of the authors in the paper. They believe that this analysis supports the case for the “introduction of a national licensing examination” in the UK.
We don’t have a national exit exam in the UK. Instead the General Medical Council (GMC) regulates medical education through individual medical schools. Quality assurance of medical education and the final exams which must be passed to gain provisional entry to the medical register rests with the GMC and external examiners.
This analysis shows that different medical schools admit students with different school qualifications, and that the higher the entrance requirements, the greater the success may subsequently be in MRCP. McManus et al refer to a study which shows this is also true of performance in MRCGP exams, but I cannot find that publication.

Discussion point 4: How should we judge the performance of medical schools? Is performance of graduates in post-graduate examinations important?
When this study was published it contributed to discussion about whether a national exit exam should be introduced. Ian Noble, a Sheffield medical student, writing in the BMJ, suggested that medical schools should be judged on whether the graduates they produced performed as competent foundation doctors, not on how well graduates performed in subsequent examinations. Since it is rare for graduates to be pulled up for poor clinical performance this suggests that we have no problem for a national exit exam to solve.

Discussion point 5: How helpful is it to read reviewer’s comments on a paper? Is this something that all journals should aim for?
This paper is published in BMC Medicine which also posts the comments of the peer-reviewers. One of the reviewers suggested that this paper should not be published as although it involved a commendable analysis of multiple datasets it did not “help me to understand the problems in medical education better, nor does it help me to improve medical education or to advance medical education as a science”. The authors’ response to this criticism is also published. The reviewers and the main author also had a dispute over another analysis published in BMC Medicine on gender and ethnicity and success in MRCP. But that discussion was pre-submission so is private correspondence between those involved.

Conflict of Interest: I’m a Belfast graduate!

Posted in Discussion Outlines, Introductions, Week 4 | 2 Comments

Week 3 – Summary: CRASH-2

This week’s journal club saw a fascinating discussion of CRASH-2 (Clinical Randomisation of an Antifibrinolytic in Significant Haemorrhage) and, as before, it has been a struggle to write a summary post that can do the discussion justice. This post will focus on the five main areas of discussion.

Study population

CRASH-2 was conducted in 40 countries world-wide and in an impressive range of healthcare settings. The criteria by which patients were deemed with or at-risk of significant haemorrhage were clinical ones. Patients were deemed to have significant haemorrhage with a systolic blood pressure less than 90 and/or a heart rate greater than 110 beats per minute. However the judgements of which patients were at risk of significant haemorrhage was purely based on clinician judgement. The question was posed by myself, ‘are these criteria robust enough?’.

The inclusion criteria were felt to be one of the strengths of the study as @ welsh_gas_doc commented

The words “at risk of…” are somewhat open to interpretation, but good inclusion criteria.

Many participants commented on the range of health care settings involved in the trial. This was felt to be very impressive as the following tweets show:


I think the clinical criteria were fine. There is an amazing range of settings involved – both developed and developing countries.


@GabrielScally Agree, the range of settings was really impressive.

Primary outcome measure

This paper reported all-cause mortality as the primary outcome to be measured. We asked whether this paper was asking the right question.

It was widely agreed that all cause mortality was the right outcome to measure, though @paramedicsa commented that:

with normal Trauma mortality at 60% within first 24hrs, shouldperiod be reduced from 28days?

…and there was an interesting discussion of whether 28 days was just an “arbitary” cut-off point, as suggested by @TWSY. @MsPhelps highlighted an interesting point – that looking at the mortality figures in the paper the all-cause mortality would probably been the same had the cut-off point been 14 days. @iamdoctord’s views did seem to reflect that of the majority of participants:

I think 28d appropriately covers the early mortality from the primary insult and secondary complications which were to be evaluated

Adverse events

No matter how effective a treatment is shown to be in any trial, it is important that any adverse effects relating to treatment are identified and reported appropriately. This paper reported no adverse events as serious, unexpected or suspected to be related to the study treatment. However as @TWSY commented the discussion of the paper stated ‘we cannot exclude the possibility of some increase in risk’. Was the reporting of adverse effects in this paper adequate?

Many felt like this issue had not been covered in enough depth in the paper: @Assidens highlighted:

really difficult to judge as all info seems to ‘play down’ possible side effects, would need more data on side effects

The paper did mention the risk of  non-fatal vascular occlusive events but the authors stated that the relative risk  of such events should be unbiased even if possibly under-reported. As @TWSY commented:

Studies are rarely large enough to detect rare events. Yes, a risk of undereporting or not recognising rare events

Tranexamic acid is now being used in clinical practice in trauma patients as a result of this trial. It is vital that there is ongoing research and monitoring of any adverse effects that may be related to this treatment in this population group as highlighted in this interesting discussion.

Mechanism of action of tranxeamic acid in trauma patients

This trial showed that tranexamic acid was effective (the number needed to treat was 67) but the precise mechanism is unknown. The question posed was should research into how this drug works in trauma patients be a major priority or is research like this showing that it does work be enough to allow its widespread use? As @iamdoctord asked:

Is full comprehension a pre-requisite for application to clinical practice

Two of the highlights of the debate surrounding question came from @welsh_gas_doc:

It would be an act of intellectual masturbation to spend money researching HOW Tranexamic Acid works. Fun, but ultimately pointless

It isn’t clear exactly how most anaesthetic drugs work; theories abound, butnobody is 100% sure . . .(I know, but it’s a secret)

The last point was one raised by several people: the precise mechanism of action of many drug treatments is unknown, paracetamol being a widely raised example, yet they are known to be clinically effective. The general consensus seemed to be that since this treatment has been shown to be effective, further research into its mechanism of action would be interesting, but should not delay this treatment’s widespread use in trauma patients with, or at risk of, significant haemorrhage.

Trials in the emergency setting – consent and delays

A subgroup analysis of the CRASH-2 trial showed that early treatment with tranexamic acid was the most effective, with treatment later than three hours actually leading to increased bleeding .One area of delay was in gaining consent for participation in the trial.  @GabrielScally pointed out that consent accounted for a delay of about an hour in administrating treatment in this trial. Roberts et al wrote a very interesting accompanying article in The Lancet about consent in emergency trials, suggesting that seeking consent may actually be unethical if it leads to a delay in the start of trial treatment, such that the effect could be reduced or obscured. The question was asked during the discussion – should consent be necessary in the emergency setting?

This lead to an extremely interesting debate (I have to admit I did play the role of devil’s advocate during this – asking “When we have real clinical uncertainty for a trial treatment should we just treat without consent?!”). @TWSY, an emergency physician, discussed how emergency care research can actually allow for retrospective consent, for example if pre-agreed by the ethics committee before the trial is run.

As with any ethical discussion opinions did vary. @TheNerdse and others felt that consent need not always be sought but this does come with caveats. From @TheNerdse:

consent need not be sought in emergency where the treatment proposed is viable, needed fast, life saving & proven. In my opinion.

But as @amcunningham pointed out this was a trial and it was conducted because the benefits of treatment were not known:

yes, but trial was conducted because benefits not known.. could have done harm… in that case consent needed

@iamdoctord raised the issue of clinical uncertainty regarding trial treatments and consent:

Despite uncertainty, and if patients know such uncertainty exists, we cannot relinquish their right to choose their therapy.

Does this apply to placebo-controlled trials where a patient’s consent allows us to allocate them to a treatment or placebo arm when we actually don’t know whether a treatment will be beneficial? This is an ethical minefield and hours could be spent discussing this one point and there would never be a consensus on how to proceed. Ultimately, trials in the emergency setting, such as CRASH-2, are very important to developing practice and can have a huge impact on how we treat future patients. Trials need to be well-planned with a robust protocol, though never forgetting that while the patient may be a participant in a trial, they are still a patient and we have ultimate responsibility for their care.

Posted in Summaries, Week 3 | Leave a comment

Week 3 – Transcript: CRASH-2

The transcript of this week’s discussion is available to view either as a PDF here or in HTML format on Twapper Keeper here.

There were mixed responses over which format people preferred, so we will be continuing with both for the time being.

Posted in Transcripts, Week 3 | Leave a comment

Week 3 – Discussion points: CRASH-2

CRASH-2 is an extremely interesting paper and I have to admit I have found it hard to narrow down the discussion points for this week’s journal club.  A huge amount of credit has to be given to @GabrielScally for turning my fairly incoherent ramblings into a sleek set of points to be used for this evening, his help has been utterly invaluable.

1. This trial had a peer reviewed protocol and the protocol seems very robust. Was the study population relevant to the question to be answered (with respect to patients deemed with or at-risk of significant haemorrhage)?

2. Did the paper ask the right question (i.e. Was the primary outcome of mortality in 28 days the right one to measure)?

3. With regards to the reporting of adverse events in the paper, do we think this was adequate to detect any adverse effects from the treatment?

4. Tranexamic acid was shown to be effective, but the precise mechanism is unclear. Should research into exactly how it works in trauma patients be a major priority, or is it enough that it does?

5. The subgroup analysis showed that the earlier the treatment is given is better – after 3 hours the treatment actually increased bleeding rates. But gaining consent for participation in the trial delayed treatment (see Roberts et al). Should consent be necessary in the emergency setting?

During the discussion tonight the abbreviation TA will be  used for tranexamic acid and I will be tweeting this before the evening’s events start at 8.00pm BST.

Also a huge thank you to @PaedsSHO and @hullodave for reading through the discussion points for me  – very much appreciated.

(For an interesting take on ethics committees and their impact on trials see this post  by @bengoldacre on his Bad Science blog and then keep going, his column Bad Science in the Guardian is a must read)

Posted in Discussion Outlines, Week 3 | Leave a comment

Week 3 – Introduction: CRASH-2

For the past two weeks Twitter journal club has looked at two fairly old but still very relevant papers. The paper chosen for this week is at the cutting edge of research and recently won the award for Research Paper of the Year at the 2011 BMJ Awards. This award recognises a paper that the judges believe will have an impact on doctors and patients over the next 5 to 10 years.

The results of CRASH-2 (Clinical Randomisation of an Antifibrinolytic in Significant Haemorrhage) were published in The Lancet in June 2010. This was a randomised placebo controlled trial which recruited 20,211 trauma patients with, or at risk of significant haemorrhage in 274 hospitals in 40 countries. Patients were randomised into two arms of the study – either to receive placebo or the drug tranexamic acid. Tranexamic acid is an antifibrinolytic, this drug works by blocking sites on an enzyme that causes blood clot breakdown. The trial was designed to assess whether early administration of this treatment (within eight hours) would have an impact on patient deaths due to bleeding.

The headline finding – all-cause mortality was significantly reduced with tranexamic acid (14.5% in the treatment group vs 16.0% in the placebo group – relative risk 0·91, 95% CI 0·85—0·97; p=0·0035). Risk of death due to bleeding was also significantly reduced (4·9% in the treatment group vs 574 5·7% in placebo group ; relative risk 0·85, 95% CI 0·76—0·96; p=0·0077).

The number needed to treat in this trial – 67, an impressive result (from all-cause mortality figures). The findings of this large scale study have the potential to have a huge impact on the management of trauma patients – the authors of the study have called for this drug to be available to doctors treating trauma patients worldwide and that it should be considered for the WHO’s List of Essential Medicines. As the CRASH 2 team said at the BMJ awards:

“Our research showed that giving tranexamic acid to bleeding trauma patients is both effective and cost effective. These findings are of global importance because every year, over 2 million people die from traumatic haemorrhage”

A very simple intervention with a relatively inexpensive drug that could have a global impact on the mortality of trauma victims worldwide – this is why we have chosen to discuss CRASH-2 in this week’s journal club.

I will be posting a list of discussion points in advance of the journal club on Sunday at 8pm BST. I am very excited to have the opportunity to critique this paper and look forward to another interesting and informative discussion.

Posted in Introductions, Week 3 | 1 Comment

Week 2 – Summary: Prevention Paradox

For our second paper we discussed Rose’s Prevention Paradox. For an introduction to this paper, please see here. I feel confident saying there was no issue with finding relevance in this paper to current practice. There was initially some concern about the choice of paper and whether the usual criteria for critiquing a paper applied, but the general consensus, both at the beginning and after the session, was that it was an interesting and important paper to discuss. I hope everyone would agree that the discussion raised some very interesting points about the nature of preventing disease, including preventive medication and public health initiatives. As with last week, we’ve picked some key points to focus on:

Impact on Clinical Practice

We began by discussing how the paper might affect clinical practice, with @petermbenglish tweeting:

This was a bright shiny new idea when the paper was published. Has it become tarnished in the  mean time?

A few people commented that this will always be an important paper, and consensus was that this is a paper relevant to many areas of clinical practice. Soon, the issue of balancing individual interests against the population’s interests came to the forefront of the discussion. @silv24 pointed out:

[We] need to examine the risks as well as the benefits when giving preventative treatment, that has to apply to every individual

and later:

Should we be aiming for improvement in the patient sat in front of us or the population as a whole?

Fundamentally, this aspect of the discussion boiled down to balancing the interests of the individual patient against the interests of the population as a whole. Most agreed that there was too much focus on high-risk patients, with @davecurtis314 suggesting this might be because they don’t understand Rose, (although in fact this can end up being in the best interests of the patient). The issue of engaging with patients to promote adherence to any medication regime was raised on a few occasions, with concerns also raised about how to encourage low-risk patients to agree to and adhere to treatment, especially in the context of treating a non-event. However, as @GarethEnticott highlighted, there need not be a dichotomy between addressing the needs of the population versus the individual:

Trouble with rose is people think it is an either/or choice. Rose concludes both strategies have to be used.

We then discussed the difference between “normalising” interventions, such as promoting healthy eating and smoking abstinence, and interventions such as medication, which effectively increase abnormality. Related to this point was the “polypill“, which most people felt would be rejected by Rose on the basis that there currently exists insufficient safety data (since safety data takes longer to obtain than efficacy data). We spoke about the fact that Rose declared long-term mass preventive medication to be unacceptable, while suggested that interventions such as promoting healthier lifestyles would be more ethical since they are “normalising” interventions. (Though @drgrumble pointed out that it may be easier to encourage patients to take a pill every day than give up smoking and change their diet.) However, as @davecurtis314 pointed out, the “problem is that effect on individual [is] tiny, effect on population/NHS huge”.

Risk Presentations

Many people felt this was something that doctors could improve on. The difference between relative and absolute risk reductions was emphasised, and number needed to treat (NNT) was also discussed, although this paper showed that presenting NNT alone to a patient may not be appropriate. This Cochrane review suggests that natural frequencies (“50 out of 1000″ rather than “5%”) may be preferable, but does not provide conclusive results. There was also a brief discussion about the teaching of statistics at medical school. The responses were mixed: @silv24 and @DrDLittle reported receiving very little teaching, though I received a lecture series in medical statistics in my first year of medical school.

Preventive vs Therapeutic: Differences between Specialities

Psychiatry was suggested by @davecurtis314 as a speciality in which prevention and treatment overlap, although prevention is usually of relapse, rather than initial episode of illness. Paediatrics was one area identified as being very pro-active in its use of preventive intervention, in the form of vaccination. A particularly interesting suggestion was made by @danjrharvey, with respect to the use of antivirals during the H1N1 influenza outbreak not only to treat ‘flu, but also to prevent transmission. Rose highlights obstetrics and antenatal care as being largely preventive by nature, both at the individual and population level. In addition, screening is particularly widely used in women’s health, and increasingly in sexual health, which can arguably count as a form of prevention. @adriamarilla also pointed out that in Cuba, doctors are trained with a much stronger population preventive approach, perhaps indicating strong geographical differences.

Most agreed that preventive treatment was more cost-effective than cures; however, as @silv24 pointed out, this raises the issue of commercial interests of pharmaceutical companies, and also the issue of public concerns about medicalisation in light of new medications with lower limits for treatment.

Social Change

The conversation they moved onto sociology: @danjrharvey felt the most powerful way of altering the behaviour of a population was through economics, and socioeconomic factors were identified as being a key predictor of health. @amcunningham tweeted:

Personally I think sociologists [are] at least as important as psychologists when thinking about what determines health behaviour.

It became apparent that social factors and lifestyle were considered very important, and @carotomes pointed out that these were both likely to be cost-effective and sustainable. There was quite a bit of discussion about whether doctors should create social pressure to effect behaviour change – is it effective? is it ethical? Social change, @amcunningham said, arose not from putting pressure on individuals, but on society as a whole, which @petermbenglish pointed out meant engaging mainstream media. No doubt doctors play a role in this too, but most agreed that this was not something they could do on their own. @aj0610, @petermbenglish and @alijmbacon raised the issue of media emphasising things that are either lacking in value, or which are wrong, with sensationalism trumping sensibility in order to appeal to what readers are interested in. However, as @anaestheticdoc pointed out:

This is presuming that society wants to change. Does everybody want good health, and go without some unhealthy lifestyle choices?

What is a Patient?

One concern with preventive treatment is that it make asymptomatic individuals assume the “sick role“. Should individuals receiving preventive treatment be considered patients? @doctorblogs quite emphatically felt “no”, and @Trisha_the_doc added that medicalising healthy people is not a good thing. Some agreed; others, such as @davecurtis314, disagreed, saying that “taking a pill every day makes you a patient”. If true, this medicalises a large proportion of the population, although as @themattmak pointed out, plenty of people take multivitamins or cod liver oil without being designated a patient. Furthermore, would we call women on the contraceptive pill “patients”?

And finally…

After the end of the meeting, the conversation continued into the evening for some time with a discussion about encouraging patients to engage in fewer unhealthy and counterproductive behaviours, such as excessive alcohol consumption and smoking.

Finally, I’d like to conclude with a quote from John Gordon, tweeted by @GabrielScally, as some food for thought:

“No mass disorder afflicting mankind has ever been brought under control or eliminated by attempts at treating the individual”


Once again, please do leave comments about this second meeting, including anything you think would help improve it. 

Posted in Summaries, Week 2 | Leave a comment

Week 2 – Transcript: Prevention Paradox

The transcript of this week’s discussion is now available: You can either view it as a PDF here: or in HTML format on a website called Twapper Keeper, here:

The PDF is, I suppose, more permanent, though the Twapper Keeper version is a little easier on the eye. If you could please comment on this post to let me know which method you prefer, I would appreciate it. Also, if you know of any other services that might be able to “thread” the @ replies and retweets, please do let us know. Twit Print is refusing to work for me, and Twitoaster seems to have closed down.

We hope you enjoyed this evening. We’ll be posting a summary of the meeting as soon as possible.

Posted in Transcripts, Week 2 | Leave a comment

Week 2 – Discussion points: Prevention Paradox

We’ve come up with an outline for how we think this evening’s discussion should be structured, based on the following points:

  1. What does this paper tell us? How relevant is this paper today? Should our priorities be for individual benefit? Or for population benefit? Where should we draw the line (i.e. have to recognise adverse side effects, cost etc)?
  2. How does the paper present risk to individuals and populations? How do we present this to patients? (Absolute vs Relative Risk Reductions and Number Needed to Treat)
  3. Has there been a separation of prevention and treatment & has this changed in last 30 years? Are some specialities more into prevention than others? Why? How can we go about integrating prevention with existing treatment? Screening programmes?
  4. Should more research be dedicated to prevention rather than cure?
  5. How do we define a patient? Should  those receiving preventive treatment for eg hypertension assume  the patient role? Is the distinction between therapeutics and preventive medicine blurred?
Based on feedback from last week, we’re going to try and keep the discussion to an hour in length (but obviously you can keep discussing afterwards using the hashtag should you wish!).

Posted in Discussion Outlines, Week 2 | 1 Comment

Week 2 – Introduction: Prevention Paradox

A few people have enquired about why I have chosen a 30 year old paper as my favourite paper to be discussed this week. Is it relevant? Surely it’s out of date? Aren’t we meant to be discussing papers that affect current clinical practice?

Despite being 30 years old, I still believe Rose’s BMJ paper is of great importance, perhaps more now than ever with widespread prescription of preventive drugs such as statins and the like. It addresses the issues of preventive medicine, presentation of risk to patients, and what strategies we can employ to best reduce the incidence of disease.


I wrote a blog in January about informed consent and preventive medicine at Occam’s Typewriter. In it, I discussed the issues of prescribing preventive medications to patients who only have a small chance of benefiting from that treatment, and how presentation of risk affects this.

Since the absolute risk reduction (and therefore likelihood of benefit from treatment) is equal to absolute risk multiplied by relative risk reduction, individuals with a low starting risk stand a smaller chance of benefiting. Consequently, when prescribing preventive medications, perhaps we should be targetting high-risk individuals who will be most likely to benefit from treatment. (I briefly explain the terms Absolute Risk Reduction (ARR), Relative Risk Reduction (RRR), and Number Needed to Treat (NNT), but a slightly more detailed explanation can be found here or here.)

The Rose paper effectively shows I am wrong. His argument is based on the fact that the number of adverse events occurring in a  group of individuals is the product of both the probability of the event occurring in a person, and the number of people in that group. Or in other words (and to quote the paper):

A large number of people exposed to a low risk is likely to produce more cases than a small number of people exposed to a high risk.

For example, while older mothers are more likely to give birth to babies with Downs syndrome, the majority of babies with Downs syndrome are born to younger mothers, simply because there are more younger mothers giving birth. Similarly, while we could restrict statin prescriptions to those with very high cholesterol, the majority of cases will arises from those with only slightly elevated cholesterol levels.

Fundamentally, the argument is between treating only those individuals at high risk, providing maximum benefit to the patient, and treating even those individuals with a much lower risk, providing the greatest benefit to the population as a whole.

Rose argues that the best way to reduce incidence of adverse events is to treat the population as a whole. Concentrating efforts on high risk individuals may be most appropriate for the patient, but its ability to reduce incidence of disease is limited.

Conversely, while treating people at lower risk levels may not benefit individuals greatly, it is the best way to reduce the occurrence of disease and illness in a population. However, Rose also argues that safety is paramount, and for individuals who stand to benefit very little from a preventive treatment, the risk of side effects may to be too great for long-term mass prevention to be acceptable.

To ameliorate this, Rose concludes with creating a distinction between “normalising behaviours”, such as healthy lifestyle choices, which can be assumed to be safe, and interventions that increase biological abnormality, such as medications. He then argues that the former may be presumed to be safe and, I assume, appropriate to promote in the population as a whole.


Another issue raised in this paper is the emphasis that is placed upon relative risk reductions. This particular sentence from the paper really stands out to me:

To express the results of trials only in terms of percentage effectiveness is to conceal what the user really needs to know.

I think this is particularly true of media reports on new drugs, both when talking about efficacy of drugs, but also risk increases as side effects associated with treatments.

For example, this paper shows that five years’ treatment with oral bisphosphonates is associated with a doubled risk of oesophageal cancer in 60-79 year olds. In fact, the risk rises from 1 in 1000 to 2 in 1000, meaning that there’s a 0.1% chance that the drug will cause you to develop one of these cancers. A risk, no doubt, but small compared to the benefits of bisphosphonates, which have been shown to reduce the risk of fractures by 30-50%. (This is particularly important for hip fractures – A meta-analysis by Haentjens et al reported a hazard ratio for all-cause deaths of 5.75 for women and 7.95 for men in the 3 months following hip fracture.)


There will no doubt be issues I have missed in this post, so please do raise them this evening. A bit before 8 o’clock (UK time, so 7pm GMT) I’ll post the key points for discussion. We are trying a different format this week, with 5 key points pertaining to the paper that we can discuss. The focus will be on how the issues raised in the paper can be applied to current clinical practice.

Posted in Introductions, Week 2 | 5 Comments