Evidence-based medicine (EBM) is a process of basing clinical practice on validated information. According to one of its modern architects, David Sackett, it is ‘the explicit, judicious and conscientious use of current best available evidence in making decisions about the care of individual patients’.^{13} According to Silagy and Haines, ‘EBM is the integration of the best available scientific evidence with your clinical expertise and knowledge, your intuition, your wisdom’.^{14}
The process of using EBM should be very comfortable for GPs because scientific methodology and evidence is second nature to us and has been the basis of our clinical decision making prior to and subsequent to graduation.
The proposed five steps of applying EBM are similar to basic research methodology:^{14}
The statistical methodologies used in EBM cover the traditional research methods but there is an emphasis on the methods of risk reduction, absolute and relative risk reduction, and number needed to treat (NNT). These definitions are included in the glossary of terms later in this chapter.
GPs have a responsibility to their patients to be well versed with the best evidence when making decisions about management (see TABLE 13.2), whether it be for a minor surgical procedure, selection of drugs, selection of an investigation, or referral to the most appropriate consultant. If the best evidence reveals that a certain practice we are using is of no value or is less efficacious than another method, then we should be prepared to change. On the other hand, if we find that a certain method works for us and there is no current evidence that it is inappropriate or the evidence is equivocal, then there is no compelling reason to change.
GPs need a healthy scepticism about what is best evidence and claims for treatment in addition to the skill of critical appraisal of research/evidence. We tend to be impressed by the perception that evidence is a numbers game. However, the great work of James Lind (see the introduction to CHAPTER 10) shows that facts do not necessarily involve large numbers.
For EBM to be accepted by GPs the information needs to be readily accessible, user friendly, significant, relevant and, perhaps, believable.
The strength of EBM is that it can provide the answers to very important everyday decisions, especially in screening and preventive medicine, where guidelines have fluctuated over the decades. The most recent RACGP guidelines for preventive activities in general practice (the red book)^{15} highlight the value of current evidence (see www.racgp.org.au/your-practice/guidelines).
GPs are currently faced with important decisions about the effectiveness of complementary therapies, which are very tempting to embrace or trial when searching for ways to manage difficult problems, such as chronic fatigue syndrome, fibromyalgia, chronic asthma, chronic pain syndromes and other difficult-to-treat diseases. We are hopeful that EBM can provide the answers to best practice in addition to evaluating individual therapies.
Remember, however, that Bayes’ theorem tells us that a positive trial result means less if the pre-existing chances of the treatment working were low to begin with. In other words, ‘extraordinary claims require extraordinary evidence’.
Glossary of terms^{18,19}
Apart from the terms and definitions used in preceding pages it is important to highlight the following terms used in EBM/research.
Absolute risk reduction (ARR) The absolute difference in event rates between two intervention or treatment groups. It gives an indication of the baseline risk and treatment effect. An ARR of 0 means no difference and thus the treatment has no effect.
Example: The ARR for prophylactic ciprofloxacin in the case cited is 10 − 2 = 8 per 100 (0.08) or 8%.
Accessing the evidence
Analysis of variance This allows comparisons between the means of two samples of similar populations with a normal distribution. The contribution to variance for each variable can be determined and tested for statistical significance.
Clinical significance Whether the benefit to people receiving an intervention compared to the control group is great enough to warrant the intervention. It is based on measure of effect.
Confidence interval A measure of the imprecision of the data results. The statistically derived range of values around a trial result in which the probability is that the true result will be within the range.
A 95% (standard) confidence interval for a sample indicates that there is a 95% chance that the interval includes the true population proportion whose circumstances comply with the evidence.
Control event rate (CER) The percentage of subjects in the control group that experienced the event of interest.
Experimental event rate (EER) The percentage of subjects in the intervention group that experienced the event of interest.
Kappa Cohen's kappa measures the agreement between the evaluations of two raters when both are rating the same object. A value of 1 indicates perfect agreement. A value of 0 indicates that agreement is no better than chance. It is an appropriate statistic for tables that have the same categories in the columns as in the rows (e.g. when measuring agreement between two raters).
Number needed to treat (NNT) The number of people who must be treated over a given period of time with the experimental therapy (specific intervention) to achieve one good outcome or prevent one adverse outcome. This incorporates the duration of treatment. It is a measure of the absolute relative risk. Obviously the lower the NNT, the better the treatment. It is calculated as 100/ARR (%); that is, the reciprocal of the ARR.
Note: The NNT will be different for different patient populations depending on their baseline risk for developing the outcome of interest.
Odds ratio The probability of the occurrence of an event compared to its non-occurrence.
Publications
Clinical evidence: BMJ Publishing Group, refer to www.clinicalevidence.org
Evidence-based medicine. BMJ Publishing Group
Probability (p) value A deceptively complex measure to understand. It is a statistical summary of the incompatibility between the observed data and what we would have expected to see if the treatment did not work in the slightest (i.e. if the ‘null hypothesis’ was true). The lower the p value, the less consistent it is that the experiment results can be explained by the null hypothesis. Confusingly, a p=0.05 does not at all equate to saying the treatment is therefore 95% likely to work. For the curious, a 2016 article offers a 14-page explanation.^{20}
Relative risk (RR) The ratio of the risk of the outcome (e.g. disease or death) in the treatment/exposure group compared with the control/unexposed group. RR informs us how many times more likely an event is to occur in the treatment group compared with the control group.
Calculation: RR = EER/CER
RR = 1 means no difference, so treatment has no effect
RR > 1 means the treatment increases the risk of disease/death
RR < 1 means the treatment decreases the risk
Example: If the risk of death from people exposed to inhalation of anthrax spores is reduced from 10 in 100 cases to 2 in 100 cases with 60 days of prophylactic ciprofloxacin, the RR of death in this group is 0.20 or 20%.
Relative risk reduction (RRR) The proportional reduction of adverse events between the treatment/experimental and the control groups in a trial (i.e. RRR is the ratio of the absolute risk reduction to the risk of the outcome in the control group). An alternative way to calculate the RRR is to subtract the RR from 1 (i.e. RRR = 1 – RR).
In the example it is: 1 − 0.2 = 0.80 or 80% or:
$RRR=ARR10=810=0.80or80%$
RRR is probably the most commonly reported measure of treatment effects, particularly when trying to emphasise the usefulness of a treatment, but the ARR gives a more realistic picture.
Risk (R) The probability that an event (death or disease) will occur.
Statistical significance The likelihood of a difference between two groups being real, based on the possibility that the difference may have occurred by chance alone. It is based on confidence intervals and p values.
Type I error A type I error occurs when a study concludes that there is a difference between two groups when there is no difference.
Type II error A type II error occurs when a study concludes that no difference exists between groups when there is a true difference.
To counterbalance the strengths (potential or real) of EBM there are concerns that it will be seized by bureaucrats to develop ‘cook book’ guidelines, Holy Writ or economic rationalisation. Others are concerned about the perceived lack of flexibility. A very interesting critical review, especially affecting psychiatry, was presented by John Ellard in his paper ‘What exactly is evidence-based medicine?’.^{16} He questioned the validity of the evidence underpinning EBM and the biases of both the proponents of ‘science’ and ‘art’ with the caution of Louis Pasteur: ‘The greatest derangement of the mind is to believe in something because one wishes it to be so’.^{17}