Plato's recordings of the dialogues of Socrates, the greatest thinker of ancient Greece, are precious items in Western philosophy, culture and science. Socratic method still informs our teaching; wise medical instructors do not fill students' heads with ‘facts’ but adopt a feigned position of ignorance from which to ask probing questions so they arrive eventually at insights that are the true prerequisite for gaining knowledge. Furthermore, these dialogues represent the birth of scientific method; starting with the recognition of one's ignorance we learn that posing pertinent questions is halfway to solving them.
In the dialogue known as ‘apology’1 (apologia – speaking in defence of one's beliefs and actions), Socrates is at his cunning best. Holding a cup of deadly hemlock, Meletus, one of his interlocutors, charges Socrates with atheism and scientific sophistry because his curiosity leads him to make enquiries into the earth and sky.
In his defence, Socrates describes how Chaerephon consulted the oracle of Delphi who confirmed that Socrates was the wisest of all men. This was a paradox to be resolved as Socrates considered himself an ignorant man. Perhaps the oracle sent Chaerephon on a divine mission to see how an ignorant man could be wiser than politicians, poets, prophets and seers? Ultimately Socrates traps Meletus into agreeing that the ignorant man who starts from the premise of knowing little is the wisest of all, thus demonstrating that posing the right questions is the secret of epistemology in all walks of life. Despite winning the argument, Socrates still had to take the hemlock.
We present a true sequence of e-mails exchanged between three clinicians contemplating the epistemology of modern medicine. Instead of sitting in the shade of an oak tree our discourse took place through the World Wide Web.
Retrospectively, the thought sequences had an eerie similarity to ‘apology’ suggesting inspiration from Ancient Greek ghosts of Socrates (JM) and Meletus (MB) overseen by Athena (SB), Goddess of Wisdom. References and clarifications (shown) were added later.
************************************
Socrates: Is Empiricism [an assertion that knowledge arises from sense experience] a way of life in the search for knowledge in the practice of medicine, or merely one useful tool among many? Are randomized controlled trials (RCTs) the only way of doing trials? Could a Bayesian approach be used sometimes?a Bookies don't start from the assumption that all horses are equal. They calculate on a system that collates the horse's overall form with how it did in its last race. Therefore, why should we, when comparing treatments, assume at the outset that all are equal?
Meletus: I don't trust Bayesian maths. This is based on prior belief systems and these are often evidence-free or even based on prejudice. I would challenge anyone to show me one example where the Bayesian approach predicted a reliable outcome to clinical research. This is different to the observation that when a medical condition has a predictable natural history then a treatment with spectacular results needs no RCT, e.g. penicillin, appendicectomy.
Athena: Have I missed something? How can Bayesian maths not be trusted? It's just maths. Is it the uses people claim for it rather than the thing itself that is problematic? My understanding was that all medical diagnosis is based on Bayesian logic. The Wizard and the Gatekeeper article helps understand this:2 there is a prior probability that your next patient with abdominal pain has appendicitis; this is different if you are a GP or a hospital surgeon after primary care filter; thus every question you ask or physical sign you elicit or even the GPs' ‘test of time’ (i.e. review in a day or next week) should add to the accuracy of the posterior probability, as per Socrates' analogy of the bookie. Diagnosis shouldn't be based on prior ‘belief’ but prior observations.
There has been a Bayesian RCT, the Growth Restriction Intervention Trial,3 based on obstetricians' prior beliefs/clinical wisdom whether growth-restricted fetuses would be better off delivered prematurely or left in utero. It worked on individual equipoise without set criteria with a combined short- (death) and long-term (disability) outcome. The stillbirth and death-before-hospital-discharge rates were equal, despite a four-day difference of gestation, but Caesareans were less frequent when waiting conservatively. Handicap was higher in <31 week group if delivered immediately. The trial was criticized (of course). You might be right that it still leaves clinicians open to their prejudices but it was quite a milestone between Bayesians and Frequentists.b It provided evidence that babies should be left in utero, if possible.
Meletus: It's not just maths, it's value laden maths with values weighted according to prior beliefs or ‘experience’. That might work reasonably with diagnosis – although one wag once put it that experience implied making the same mistakes again. When it comes to treatment of conditions with an unknown or unpredictable natural history then, as far as I'm concerned, Bayesian approaches are simply inductive logic, selecting the evidence that reinforces your prejudices. If we had used a Bayesian approach instead of the deductive approach of RCTs we would still be doing radical mastectomy as the prior belief was set so far away to the left of the null hypothesis that no amount of data could shift it into the non-inferiority domain of breast conservation surgery.4
Socrates: But once we have RCT results, don't we have to use inductive logic if we are to apply them in practice? There must be a degree of generalization from the particular result. Or should we only believe a drug is effective if our patient was involved in the RCT? Does medicine in practice cease to be scientific because it uses such induction? Am I misunderstanding what you mean by induction?
Meletus: I attach my Karl Popper memorial lecture in my defence ( Box 1)!5
Box 1. The philosophical surgeon: in defence of evidence-based medicine.5 Precis of Karl Popper Memorial Lecture, London School of Economics, 2007.
Aristotle and other great names of the golden age of Pericles in the ancient city of Athens were the first to apply a systematic approach to the pursuit of knowledge. They considered that our beliefs of the world around us were figments of our imagination, and it was, therefore, necessary to systematically collect observations to challenge these views. These observations were built up into a conceptual model (hypothesis), and later observations were selected to corroborate this model.
The process of collecting observations in defence of a hypothesis is known as inductivism. Inductive logic was considered ‘science’ up until the 18th century, when the Scottish philosopher David Hume finally illustrated the poverty of the process. Perhaps the best way of illustrating this as it relates to our lives as medical practitioners is to consider the subject of alternative medicine.
When doctors attack alternative medicine or appear sceptical to its much trumpeted claims, we are often accused of being bigots with closed minds, protecting a closed shop. In fact the opposite is true. The alternative to alternative medicine should be scientific medicine, not ‘orthodoxy’. By science, I mean the application of deductive logic. The deductive approach starts with the formulation of the hypothesis, but for a start the hypothesis must be rational in its explanation of the disease process or therapeutic intervention. By ‘rational’ I mean built upon the growth of knowledge of human biology and physiology from the past 100 years or so, without invoking magic or metaphysical principles.
I illustrate these points with history from my own specialist field – breast cancer. Galen believed that breast cancer was due to an excess of black bile (melancholia). Inductive support for this belief came from the observation that breast cancer was more common in postmenopausal women than premenopausal women; this was thought to be because the menstrual flux in premenopausal women removed the putative excess of black bile. Therefore, the therapeutic consequences of this belief were purgation and venesection (bloodletting).
More recently, it was long assumed that radical mastectomy was the best treatment for breast cancer on the basis of anecdote. However, in the 1960s Dr Bernard Fisher proposed an hypothesis which suggested that this is false and that breast cancer is normally already a systemic disease at the time of diagnosis. The consequences are that the best treatments are now local control and systemic therapies such as chemotherapy and hormonal treatments which have dramatically improved life expectancy for women diagnosed with this dread disease. Dr Fisher who first proposed this hypothesis, used randomized controlled trial evidence to explore and challenge this revolutionary set of ideas.
[After a pause]
Socrates: I would like to respond to your excellent lecture. I would distinguish between ‘inductivism’ (an entire view of scientific methodology) which I agree is thoroughly discredited, and ‘induction’ itself, which is ‘any inference where the claim made by the conclusion goes beyond the claim jointly made by the premises’.6 If we throw out induction with ‘inductivism’ we lose something we all use daily in medical practice (I will argue). Hume explained this very clearly. Induction, in his argument, uses an empirical experience (a past event by definition) as the foundation to predict future events. He says there is no empirical foundation for doing this. He acknowledges that we all do it. He does it. But there is no foundation for it.
‘It is impossible, therefore, that any arguments from experience can prove this resemblance of the past to the future, since all these arguments are found on the supposition of that resemblance. Let the course of things be allowed hitherto ever so regular, that alone, without some new argument or inference, proves not that for the future it will continue so. In vain do you pretend to have learned the nature of bodies from your past experience. Their secret nature, and consequently all their effects and influence, may change without any change in their sensible qualities. This happens sometimes, and with regard to some objects. Why may it not happen always, and with regard to all objects? What logic, what process of argument secures you against this supposition? My practice, you say, refutes my doubts. But you mistake the purport of my question. As an agent, I am quite satisfied in the point; but as a philosopher who has some share of curiosity, I will not say scepticism, I want to learn the foundation of this inference.’7
Athena: And how does this relate to modern medical practice?
Socrates: Applying RCT evidence to the patient in front of me is exactly this sort of inductive move Hume says is not justified on the basis of empirical knowledge. This is the classic problem of induction (as I understand it). Inductive inferences assume what has been called the ‘uniformity of nature’: that the future will resemble the past. This is a fundamental philosophical assumption of science. The future is not based on empirical observation – by definition!
I have put all this together in a logical form.
Consider the following deductive arguments:
- Using past evidence to predict future events is an inductive process
- Medical practice uses RCT observations and applies them to patients in the present/future
- Conclusion: Medical practice is based on induction
- Inductive logic is unscientific
- Medical practice is based on induction
- Conclusion: Medical practice is unscientific
Because they are deductive arguments, in order to disagree with the conclusions you would have to disagree with one or other of the premises (assuming I have framed the arguments correctly)…
I would disagree with the premise 2 (a) – that inductive logic is unscientific. What would you say?
Meletus: A good challenge. I think frequentists in clinical methodology accept this argument in part and have been trying to address it for some time indirectly through power calculations. Strictly, it is true that the result of an RCT applies only to the population studied. The more prescriptive the entry criteria the more difficult it is to generalize beyond the trial population.
Because of this I belong to the pragmatic wing of the evidence-based medicine (EBM) movement. We believe that the larger the sample size and the broader the entry criteria and the intention-to-treat analysis the more unlikely it becomes that the result can't be applied to the patient in front of you …
Athena: Too many double negatives!
Meletus: … It becomes statistically improbable that your patient will respond in a different way – that is beyond the 95% confidence interval of the trial result. Furthermore, if there are such outliers we are duty bound to explain them. In other words, if there is a defined subgroup who demonstrate statistical heterogeneity from the main group, we can learn from them. I've described this as ‘biological fall-out’. The larger the sample size, the more we can look at outliers and the more we can learn about the disease and its treatment. That is why large collaborative groups try to collect tens of thousands of patients or carry out meta-analysis, not out of megalomania, nor just to find tiny incremental improvements for the whole group, but ultimately to personalize care. Eventually this allows the patient and doctor to make their own knowledge-based informed choice. It's a little better than ‘ignorance-based medicine’.
Socrates: A fine clinician's answer with good research backing, which is pragmatic and sensible and something to aspire to. But it does not avoid the charge of induction. It is simply an inductive argument which attempts to increase the support of the truth of the conclusion to a higher degree, but does not and cannot guarantee the truth of the conclusion. The conclusion is not a deductive conclusion flowing inescapably from the premises.
Popper, I think, would say that the only amount of powering that gives empirical justification is infinite powering. Anything less is finite, and a finite number is always infinitely small compared to infinity. So, the problem of induction remains because there is no certain ‘medical proof’ based on experience. I agree with every word you say, as long as you admit it is a process of induction!
Meletus: My dictionary has nine meanings for induction so we are in danger of generating a futile semantic exercise. By chance, ‘inductile’ (‘not pliable; unyielding to influences’) appears above ‘induction’, a word I've never used but appears apposite. I suggest that RCTs are exercises in the method of the hypothetico-deductive cascade that is the never-ending, constantly refined approximation to an objective reality. The treatments that are the therapeutic consequences of the biological model that best represents our transient approximation to reality are, therefore, the best available at the time. Our practice, however, is NOT ‘inductile’ as it will yield to treatments that emerge as we improve our approximation to a truth that is always beyond the reach of us mere mortals.
Socrates: I hope you agree it's worth putting the discussion in some context. It is interesting that our views will seem to converge almost completely by the end. I agree definitions are complicated. I was, however, trying to use what I understand to be the mainstream philosophical meaning of induction: ‘any inference where the claim made by the conclusion goes beyond the claim jointly made by the premises’. You might be interested in a quote from Popper which relates to our discussion about powering:
‘One (could) ascribe to the hypothesis a certain probability … on the basis of an estimate of the ratio of all tests passed by it to all the tests which have not yet been attempted … This estimate can, as it happens, be computed with precision, and the result is always that the probability is zero.’8
In other words, no amount of positive empirical data can ever raise the probability of a scientific theory above zero. Now is the time to question if you are a thorough going Popperian! On the basis of experience alone, however, it would seem that Popper is right (as most philosophers accept) – there is no purely empirical foundation for scientific theories.
Meletus: The problem is clear.
Socrates: If we step back, we can see why this is worth discussing. The questions seem to be ‘Is there such a thing as a systematic approach to knowledge that distinguishes science from pseudo-science? If there is, what is it? If there isn’t, is there anything useful that can be said to make a distinction?'
Popper's rejection of induction is part of the problem – not part of the solution I think. Popper successfully critiques the verification principle of logical positivism. Using empirical data and logic alone we cannot ‘verify’ a scientific theory. He thus rejects induction, rejects verification and opts for falsification (whereby a true scientific theory is identified as one which is falsifiable in principle).
If we follow the skepticism (or curiosity) of Hume and Popper, then we have to conclude that we cannot find firm foundations in science based on empirical justification alone.
Athena: What about actual scientists?
Socrates: For these reasons I am not a Popperian (is that heresy?), and neither are most practising scientists , while agreeing that modern science still rests on his insights as much as any other individual. Since Popper – and probably because of Popper's ‘anti-foundationalist’ insights – many philosophers of science have given up on a ‘systematic approach to knowledge’. Practising scientists do seem convinced to a high degree of the truth of a particular theory, which is more than Popper seems to allow for. There is no singular scientific method, whether logical positivism or falsification. There are instead a variety of methods used in different degrees in different contexts.
Meletus: But surely we are left without either being able to define ‘true science’ or to identify and reject ‘pseudo-science’? Homeopaths will always be able to claim, as some do, that RCTs are not the appropriate tool for investigating homeopathy.
Socrates: It is, I concede, more challenging for those of us wishing to promote public understanding of science-based medicine. This culture is philosophically justified in being sceptical of statements regarding science which speak in terms of ‘facts’, ‘proofs’ and ‘certainties’. Popper has successfully refuted this approach to science.
Scientific epistemology is complex. While the place of empiricism is perhaps weaker than we might like, it would be a mistake to try to find ‘certainty’ somewhere else. Empirical observation remains essential, not merely important, for good science. Philosophically speaking, it is a necessary part of science, but it is not sufficient for good science. There will always be assumptions (such as the uniformity of nature) which themselves are not empirically verifiable but which we need to make in order to do good science.
Athena: I feel a bit weak …
Socrates: Don't. One response I favour is called ‘Critical Realism’. There is a real world out there, real things can be known about it, and we need to be highly critical in our epistemology, always accepting that we might be wrong, but also believing that it is possible to be right (just not with certainty). What critical realism looks like in detail will vary from discipline to discipline. Rather than signing up to a particular scientific method as ‘the right way’, this is described as a humble epistemology. Far from giving licence to justify pseudo-science, critical realists constantly work to submit knowledge claims to a higher authority than personal whims and prejudices – reality itself.
To ‘understand’ is literally to ‘stand under’. Too often in history people have seen ‘knowledge’ as a conquest of reality rather than a submission to it. Instead of recognizing true science by possession of correct methodology, it might be better to identify a moral element; the virtue of humility. By this standard ‘pseudo-science’ is best identified as arrogantly failing to submit ‘knowledge’ claims to reality.
I, therefore, think that the aims of HealthWatchc would be best achieved with a gentle and humble confidence rather than speaking of certainties.9 I realize that scientists who are generally more ‘modernist’ by temperament find this approach a little namby-pamby. However, I think tentative statements are actually more persuasive than absolutist statements – we can leave such ‘certainties’ to the hucksters and fraudsters, while being confident that they are wrong. This seems to me to be exactly what you expressed much more concisely than me in an earlier e-mail. A ‘cascade that is the never ending, constantly refined approximation to an objective reality’ – but please consider that it may not be as Popperian as it first appears, and the ‘hypothetico-deductive cascade’ is not the only game in town.
Meletus: I declare that I am no longer a Popperian but a Critical Realist. Philosophers have the advantage over clinicians that their belle penses cannot kill people first hand although, as Popper prophesied,10 extreme ideologies from the left and the right will kill millions. Nevertheless, we clinicians have to make decisions on a daily basis in the face of uncertainty and learn from our mistakes. This is an exercise in humility whereas ‘experience’ (in the sense of ‘the way things have always been done’) simply repeats past mistakes. As you describe so eloquently, a good clinician must have many qualities above and beyond technical skills. Among these are humility, scepticism and a willingness to subject favoured ideas to the hazard of refutation. It seems that we have closed the gap between our starting positions.
Athena: Gentlemen, it appears that Socrates has won the argument. I wonder if we should share what we have learned: that we must hold our medical knowledge lightly and with humility, with confidence rather than certainty, and that we need constantly to be asking questions of our preferred epistemological theories?
Footnotes
DECLARATIONS —
Competing interests None declared. Our only worry is that around the time of this interchange, JM (Socrates) was elected chairman of the UK charity Healthwatch, which may yet prove to be his poisoned chalice!
Funding None
Ethical approval Not applicable
Guarantor JM
Contributorship All authors contributed equally
Acknowledgements
We thank Caroline Richmond for the remark that set off this discussion and Mandy Payne for helpful comments on the manuscript
Footnotes
Bayes' theorem interprets the concept of probability as a measure of a state of knowledge in contrast to interpreting it as a frequency or a physical property of a system
Frequentism: a statistical perspective that focuses on the frequency with which an observed value is expected in numerous trials in an effort to avoid anything savouring of matters of opinion
Healthwatch-UK is a charity that promotes evidence-based medicine
References
- 1.Benson Hugh H. Socratic Wisdom: The Model of Knowledge in Plato's Early Dialogues. New York, NY: Oxford University Press; 2000 [Google Scholar]
- 2.Mathers N, Hodgkin P. The Gatekeeper and the Wizard: a fairy tale. BMJ 1989;298:172–4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.GRIT Study Group A randomised trial of timed delivery for the compromised preterm fetus: short term outcomes and Bayesian interpretation. BJOG 2003;110:27–32 [DOI] [PubMed] [Google Scholar]
- 4.Fisher B, Jeong J-H, Anderson S, Bryant J, Fisher ER, Wolmark N. Twenty five-year follow-up of a randomized trial comparing radical mastectomy, total mastectomy,and total mastectomy followed by irradiation. N Engl J Med 2002;347:567–75 [DOI] [PubMed] [Google Scholar]
- 5.Baum M. Karl Popper Memorial. Lecture: The philosophical surgeon: in defence of evidence-based medicine. London: London School of Economics; 2007. See www2.lse.ac.uk/PublicEvents/pdf/20071106_Popper.pdf [Google Scholar]
- 6.Audi R, ed. Cambridge Dictionary of Philosophy. 2nd edn Cambridge: Cambridge University Press; 1999 [Google Scholar]
- 7.Hume D. An Inquiry Concerning Human Understanding. Indianapolis: [publisher unknown]; 1748 [Google Scholar]
- 8.Popper KR. The Logic of Scientific Discovery. Abingdon: Routledge Classics; 2002 [Google Scholar]
- 9.May J. The possibility of knowledge in a post-modern world. HealthWatch Newsletter 67. London; HealthWatch; 2007. See http://www.healthwatch-uk.org/newsletterarchive/nlett67.pdf [Google Scholar]
- 10.Popper KR. The Open Society and Its Enemies. Volume 2: The Tide of Prophecy Aftermath. London: Routledge; 1992 [Google Scholar]