How to Read a Paper: The Basics of Evidence-Based Medicine (5 page)

Experienced clinicians might think they can answer Mrs Jones' question from their own personal experience. As I argued in the previous section, few of them would be right. And even if they were right on this occasion, they would still need an overall system for converting the rag-bag of information about a patient (an ill-defined set of symptoms, physical signs, test results and knowledge of what happened to this patient or a similar patient last time), the particular values and preferences (utilities) of the patient and other things that could be relevant (a hunch, a half-remembered article, the opinion of a more experienced colleague or a paragraph discovered by chance while flicking through a textbook) into a succinct summary of what the problem is and what specific additional items of information we need to solve that problem.

Sackett and colleagues, in a book subsequently revised by Straus [15], have helped us by dissecting the parts of a good clinical question:

• First, define precisely
  whom
  the question is about (i.e. ask ‘How would I describe a group of patients similar to this one?’).
  
• Next, define
  which
  manoeuvre you are considering in this patient or population (e.g. a drug treatment), and, if necessary, a comparison manoeuvre (e.g. placebo or current standard therapy).
  
• Finally, define the desired (or undesired)
  outcome
  (e.g. reduced mortality, better quality of life, and overall cost savings to the health service).
  

The second step may not concern a drug treatment, surgical operation or other intervention. The manoeuvre could, for example, be the exposure to a putative carcinogen (something that might cause cancer) or the detection of a particular surrogate endpoint in a blood test or other investigation. (A surrogate endpoint, as section ‘Surrogate endpoints’ explains, is something that predicts, or is said to predict, the later development or progression of disease. In reality, there are very few tests that reliably act as crystal balls for patients' medical future. The statement ‘The doctor looked at the test results and told me I had six months to live’ usually reflects either poor memory or irresponsible doctoring!) In both these cases, the ‘outcome’ would be the development of cancer (or some other disease) several years later. In most clinical problems with individual patients, however, the ‘manoeuvre’ consists of a specific intervention initiated by a health professional.

Thus, in Mrs Jones's case, we might ask, ‘In a 68-year-old white woman with essential (i.e. common or garden) hypertension (high blood pressure), no coexisting illness, and no significant past medical history, whose blood pressure is currently X/Y, do the benefits of continuing therapy with bendroflumethiazide (chiefly, reduced risk of stroke) outweigh the inconvenience?’. Note that in framing the specific question, we have already established that Mrs Jones has never had a heart attack, stroke or early warning signs such as transient paralysis or loss of vision. If she had, her risk of subsequent stroke would be much higher and we would, rightly, load the risk–benefit equation to reflect this.

In order to answer the question we have posed, we must determine not just the risk of stroke in untreated hypertension but also the likely reduction in that risk which we can expect with drug treatment. This is, in fact, a rephrasing of a more general question (do the benefits of treatment in this case outweigh the risks?) which we should have asked before we prescribed bendroflumethiazide to Mrs Jones in the first place, and which all doctors should, of course, ask themselves every time they reach for their prescription pad.

Remember that Mrs Jones' alternative to staying on this particular drug is not necessarily to take no drugs at all; there may be other drugs with equivalent efficacy but less disabling side effects (as Chapter 6 argues, too many clinical trials of new drugs compare the product with placebo rather than with the best available alternative), or non-medical treatments such as exercise, salt restriction, homeopathy or yoga. Not all of these approaches would help Mrs Jones or be acceptable to her, but it would be quite appropriate to seek evidence as to
whether
they might help her—especially if she was asking to try one or more of these remedies.

We will probably find answers to some of these questions in the medical literature, and Chapter 2 describes how to search for relevant papers once you have formulated the problem. But before you start, give one last thought to your patient with high blood pressure. In order to determine her personal priorities (how does she value a 10% reduction in her risk of stroke in 5 years' time compared to the inability to go shopping unaccompanied today?), you will need to approach Mrs Jones, not a blood pressure specialist or the Medline database! Chapter 16 sets out some structured approaches for doing this.

Exercise 1

1.
Go back to the fourth paragraph in this chapter, where examples of clinical questions are given. Decide whether each of these is a properly focused question in terms of

a.
the patient or problem;

b.
the manoeuvre (intervention, prognostic marker, exposure);

c.
the comparison manoeuvre, if appropriate;

d.
the clinical outcome.

2.
Now try the following:

a.
A 5-year-old child has been on high-dose topical steroids for severe eczema since the age of 20 months. The mother believes that the steroids are stunting the child's growth, and wishes to change to homeopathic treatment. What information does the dermatologist need to decide (i) whether she is right about the topical steroids and (ii) whether homeopathic treatment will help this child?

b.
A woman who is 9 weeks pregnant calls out her general practitioner (GP) because of abdominal pain and bleeding. A previous ultrasound scan showed that the pregnancy was not ectopic. The GP decides that she might be having a miscarriage and tells her she must go into hospital for a scan and, possibly, an operation to clear out the womb. The woman is reluctant. What information do they both need in order to establish whether hospital admission is medically necessary?

c.
A 48-year-old man presents to a private physician complaining of low back pain. The physician administers an injection of corticosteroid. Sadly, the man develops fungal meningitis and dies. What information is needed to determine both the benefits and the potential harms of steroid injections in low back pain, in order to advise patients on the risk–benefit balance?

References

1
Sackett DL, Rosenberg WM, Gray J, et al. Evidence based medicine: what it is and what it isn't.
BMJ: British Medical Journal
1996;
312
(7023):71.

2
Sackett DL, Haynes RB. On the need for evidence-based medicine.
Evidence Based Medicine
1995;
1
(1):4–5.

3
Greenhalgh T. Is my practice evidence-based?
BMJ: British Medical Journal
1996;
313
(7063):957.

4
Dubinsky M, Ferguson JH. Analysis of the National Institutes of Health Medicare coverage assessment.
International Journal of Technology Assessment in Health Care
1990;
6
(03):480–8.

5
Sackett D, Ellis J, Mulligan I, et al. Inpatient general medicine is evidence based.
The Lancet
1995;
346
(8972):407–10.

6
Runciman WB, Hunt TD, Hannaford NA, et al. CareTrack: assessing the appropriateness of health care delivery in Australia.
Medical Journal of Australia
2012;
197
(10):549.

7
Macnaughton J. Anecdote in clinical practice. In: Greenhalgh T, Hurwitz B, eds.
Narrative based medicine: dialogue and discourse in clinical practice
. London: BMJ Publications, 1998.

8
Greenhalgh T. Narrative based medicine: narrative based medicine in an evidence based world.
BMJ: British Medical Journal
1999;
318
(7179):323.

9
Greenhalgh T. Intuition and evidence—uneasy bedfellows?
The British Journal of General Practice
2002;
52
(478):395.

10
Mori R, Lakhanpaul M, Verrier-Jones K. Guidelines: diagnosis and management of urinary tract infection in children: summary of NICE guidance.
BMJ: British Medical Journal
2007;
335
(7616):395.

11
Mulrow CD. Rationale for systematic reviews.
BMJ: British Medical Journal
1994;
309
(6954):597.

12
van Helmont JA.
Oriatrike, or physick refined: the common errors therein refuted and the whole art reformed and rectified
. London: Lodowick-Loyd, 1662.

13
Swinglehurst DA. Information needs of United Kingdom primary care clinicians.
Health Information & Libraries Journal
2005;
22
(3):196–204.

14
Smith R. Where is the wisdom…?
BMJ: British Medical Journal
1991;
303
(6806):798.

15
Straus SE, Richardson WS, Glasziou P, et al.
Evidence-based medicine: how to practice and teach EBM
(Fourth Edition). Edinburgh: Churchill Livingstone, 2010.

Chapter 2

Searching the literature

Evidence is accumulating faster than ever before, and staying current is essential for quality patient care.

Studies and reviews of studies of doctors' information-seeking behaviour confirm that textbooks and personal contacts continue to be the most favoured sources for clinical information, followed by journal articles (see, e.g. [1]). Use of the Internet as an information resource has increased dramatically in recent years, especially via PubMed/Medline, but the sophistication of searching and the efficiency in finding answers has not grown apace. Indeed, ask any medical librarian and you will hear tales of important clinical questions being addressed using unsystematic Google searches. While the need of healthcare professionals for information of the best quality has never been greater, barriers abound: lack of time, lack of facilities, lack of searching skills, lack of motivation and (perhaps worst of all) information overload [2].

The medical literature is far more of a jungle today than it was when the first edition of this book was published in 1996. The volume and complexity of published literature has grown: Medline alone has over 20 million references. While Medline is the flagship database for journal articles in the health sciences, it is a very conservative resource, slow to pick up new journals or journals published outside the USA, so there are many thousands of high-quality papers that may be available via other databases but are not included in Medline's 20 million. The proliferation of databases makes the information jungle that much more confusing, especially because each database covers its own range of journals and each has its own particular search protocols. How will you cope?

There is hope: in the past decade, the information ‘jungle’ has been tamed by means of information highways and high-speed transit systems. Knowing how to access these navigational wonders will save you time and improve your ability to find the best evidence. The purpose of this chapter is not to teach you to become an expert searcher but rather to help you recognise the kinds of resources that are available, choose intelligently among them and put them to work directly.

What are you looking for?

A searcher may approach medical (and, more broadly, health science) literature for three broad purposes:

• Informally, almost recreationally, browsing to keep current and to satisfy our intrinsic curiosity;
  
• Focused, looking for answers, perhaps related to questions that have occurred in clinic or that arise from individual patients and their questions;
  
• Surveying the existing literature, perhaps before embarking on a research project.
  

Each approach involves searching in a very different way.

Browsing
has an element of serendipity about it. In the old days, we would pick up our favourite journal and follow where our fancy took us. And if our fancy was informed with a few tools to help us discriminate the quality of papers we found, so much the better. These days, we can make use of new tools to help us with our browsing. We can browse electronic journals just as easily as paper journals; we can use alerting services to let us know when a new issue has been published and even tell us if articles matching our interest profile are in that issue. We can have Rich Site Summary (RSS) feeds of articles from particular journals or on particular topics sent to our e-mail addresses or our i-Phones or personal blogs, and we can participate in Twitter exchanges related to newly published papers. Almost every journal has links from its home page allowing at least one of these social networking services. These technologies are changing continuously. Those of us who have been faced with deluges of new off-prints, photocopies and journal issues we have been meaning to read will be happy to learn that we can create the same chaos electronically. That is what browsing serendipitously is all about, and it is a joy we should never lose, in whatever medium our literature may be published.