One of the major changes in the teaching and practice of medicine has been the rapid growth in evidence-based medicine (EBM), which is reflected in this new edition by now having its own subsection. As Paul Glasziou and colleagues have argued ‘a 21st century clinician who cannot critically read a study is as unprepared as one who cannot take blood pressure or examine the cardiovascular system’ (Glasziou et al., ). Evidence-based medicine, previously referred to as clinical epidemiology, has grown rapidly over the last 20 years, partially as a result of better-quality research, systematic methods to accumulate and summarise these data, and easy access to high-quality databases such as the Cochrane collaboration or EBM-based guidelines that allow health professionals to quickly access evidence when considering patient management.

What is evidence-based medicine?

There are several different definitions of EBM but we favour a modified version of what was proposed by David Sackett and colleagues as ‘the conscientious, explicit and judicious use of current best evidence in making decisions about the care of individual patients and improving the health of populations’ (words in italics have been added by us to incorporate a broader public health dimension). EBM uses the hierarchy of evidence (see Chapter 5, Box 5.1) to help weigh up the relative importance of different types of evidence. It is common today to usually use evidence from more than one study by systematically collating and synthesising data across all studies using the technique of meta-analysis (see Chapter 12 for more details). In general, randomised trials are placed above clinical experience but EBM does not ignore such experience especially in the absence of good-quality trials. When applying the hierarchy of evidence it is important to appreciate that the ordering only applies if the evidence from each type of design comes from well-conducted studies as otherwise this may be misleading. For example, a poorly conducted trial may be less valid than data from a high-quality cohort study.

This definition of EBM counters some common misperceptions about EBM highlighted in Box 8.1. Listening to patients’ preferences is an essential aspect of good medical practice and these can be formally used in EBM by allocated scores (utilities) to different outcomes or side effects so that a treatment may appear to have more benefit than harm in one patient than another (see Chapter 13 on health economics). This prevents a cookbook approach and at the end of the day the health professional may choose to follow a different management path than that promoted by a guideline, though they should be able to justify this, especially if this goes against national or international evidence-based guidelines. For example the National Institute for Health and Clinical Excellence in March 2012 had guidance covering 363 interventional procedures, 249 technology appraisals, 147 clinical guidelines and 36 Public Health guidance documents. It also provides an evidence-based portal through its own website NHS Evidence (https://www.evidence.nhs.uk/).

One of the biggest criticisms about practising EBM is that it is simply not practical in a busy clinic. This is a valid point though electronic access to high-quality summary reports (e.g. critically appraised topics CATs) is making this less relevant and enthusiasts such as Paul Glasziou have demonstrated how they can practice EBM in ‘real time’ as long as one has internet access and access to key databases and journals.

One of the key terms in the above definition is the word explicit. This means that EBM compared to non-EBM practitioners should use quantitative data to help them and patients decide on management options. Whilst both types of practitioners may recommend the same drug treatment, the EBM doctor is more likely to explain the harms and benefits by using information on the number needed to treat or the probability or risk of an event. This probabilistic way of thinking can be difficult for some doctors (see Chapter 7) as it describes an average effect which may not reflect the experience of the patient in front of them. Variation (also known as heterogeneity) in the potential benefits of treatment has been used as a criticism of EBM, but it is possible to take this into account to some degree by translating the relative benefits of treatment (e.g. relative reduction in mortality 20%) into absolute benefits taking into account individual patient characteristics. For example, after a myocardial infarction, the benefits of aspirin in terms of preventing future cardiovascular events outweigh any harm from bleeding side effects. This is because the absolute risk of another cardiovascular event is high in this population. However, in a younger healthier patient with a much lower absolute risk, the harm from aspirin may now outweigh the potential benefits, even if in relative terms the benefits are the same. Advice concerning aspirin is constantly under review as new evidence emerges. For example, recent research suggests that aspirin may now also reduce the risk of cancer and metastatic spread.

The EBM Domains

Any clinical encounter can result in a number of issues that are amenable to an EBM approach. These are grouped into the five domains as listed in Table 8.1. These will all be covered in the subsequent chapters other than aetiology which we have addressed in the epidemiology section.

Table 8.1 EMB domains.

For example on a paediatric ward round you may encounter a child who has recently been admitted with their first episode of shortness of birth which was subsequently diagnosed as asthma. One could consider the following questions: what is the most useful diagnostic test (or set of tests) to differentiate asthma in a child from other causes of shortness of breath? Will this child have persistent asthma symptoms in adulthood and will this have any long term functional limitations? Did exposure to allergens or chemicals in childhood have a role in their development of asthma? Will maintenance of inhaled steroids reduce the likelihood of future admissions and are they cost-effective?

What is the evidence that EBM changes the way we practise?

Paradoxically, the evidence behind the benefits of EBM is limited mainly because it is very hard to undertake research to evaluate EBM teaching. A systematic review (see Chapter 11) undertaken in 2004, on postgraduate training found 23 articles but most were either before and after comparisons, or nonrandomised comparison studies with only two being randomised controlled trials. Not surprisingly, most assessed knowledge and found that some form of teaching (workshop, journal club, seminars) improved understanding and knowledge and critical appraisal skills. Some studies, but not all, found that those taught EBM had more positive attitudes to EBM and behaviour change but when this was integrated with clinical practice. None of the studies measured changes in health outcomes but they were not designed to do this.

Stages of EBM

There are several steps in being a EBM practitioner. These are (a) formulate a clear question, (b) search for the evidence, (c) critically appraise the evidence, (d) apply the evidence (or not) to the individual patient or population as appropriate.

Sackett coined the acronym PICO (Patient, Intervention, Comparator, Outcome) as a helpful tool in formulating EBM questions. For aetiological and prognostic questions the intervention is equivalent to the risk or prognostic factor and for diagnostic questions this would be a diagnostic test. Table 8.2 gives two examples.

Table 8.2 Two examples of EBM questions.

The process of finding evidence is becoming simpler as more databases are being developed that synthesize the evidence though it may be necessary to go back and read the original primary studies. To undertake a full bibliographic database search (e.g. Medline) can be very time-consuming and will identify many irrelevant papers as well as potentially missing some important ones depending on the quality of the search strategy. Databases such as the Cochrane collaboration (http://www.cochrane.org/) are extremely valuable though are mainly limited to intervention studies. In the United Kingdom, the NHS evidence website (https://www.evidence.nhs.uk/) provides a portal to other sources of evidence including EBM guidelines as produced by the National Institute for Health and Clinical Excellence (http://www.nice.org.uk/). Other databases are focussed on certain disciplines or settings such as BestBets (http://www.bestbets.org/) that was initially designed for emergency room or on-call problems (http://www.eboncall.org/).

Having found some evidence, it will be necessary to appraise it to some degree. The subsequent chapters should provide you the knowledge and skills to undertake such an appraisal but like most things you only get good by constant practice. If one is fortunate, then there will already be a published high-quality meta-analysis or highly respected guideline that cites the evidence. It should be noted that much ‘evidence’ is based on expert or consensus opinion which may or may not be evidence-based and even in the presence of evidence can be biased. For example a comparison of 24 Cochrane reviews with industry sponsored or other reviews found that the industry sponsored reviews were more likely to recommend the therapy and with fewer reservations than the Cochrane reviews even though the cited evidence was essentially the same (Jorgensen et al.,).

Different groups of experts may produce different guidelines and this may reflect cultural attitudes or financial incentives to patient management. Even apparently neutral evidence should be read with some caution. For example a review of 53 Cochrane reviews highlighted some problems with the reviews in almost a third with about a fifth where the authors felt that the conclusions of the review were not supported by the evidence (Olsen et al., ). There are now a myriad of methodological checklists (see http://www.unisa.edu.au/cahe/CAHECATS/ such as CONSORT (single trials), STROBE (observational studies), PRISMA (systematic review and meta-analysis of trials) MOOSE (meta-analysis of observational studies), STREGA (genetic association studies. Such checklists are a useful tool to remind you of what key aspects of design, analysis and interpretation can go wrong but should not be applied blindly as there is a danger of rejecting evidence on the basis of some methodological problem even when this may mean the real benefit could actually be even more substantial.

Finally there is the issue of generalisability and whether the evidence is relevant to the individual patient that generated the question in the first place. There will always be an element of subjectivity in such a decision and in some cases one will need to apply external knowledge such as pharmacology to decide whether the findings in one population should or should not apply to another. For example, many therapeutic trials do not include many ethnic minority patients so a benefit seen in a Caucasian population may or may not apply to South Asian patients. In this case a judgement has to be made about whether it is reasonable or not to generalise such findings and a discussion about the uncertainty with the patient should be undertaken.

Often an EBM search can highlight the absence of evidence or only the presence of poor quality evidence. For example a review of 109 inpatients seen in Oxford for one month found that for 53% of primary treatments there was trial evidence to support therapy. In an additional 29% there was convincing nonexperimental evidence and in 18% there was no evidence that therapy was better than no therapy (Ellis et al.,). This figure is likely to be less good for some other specialties e.g. primary care. Such absence of evidence does not mean evidence of absence and should act as a stimulus for future research to help fill such evidence-based gaps.

Stay updated, free articles. Join our Telegram channel

Join