Secondary Logo

Journal Logo

Evidence-Based Medicine

Sharma, Sanjay MD, MSc (epid), MBA


Traditionally, the practice of medicine has been based on principles which are deeply rooted in pathophysiology, deductive reasoning, and clinical experience. However, we are entering into a new era of medicine in which patients, clinicians, and governments, in part fueled by their unprecedented access to medical information, are demanding that medical practice no longer be based on hypothetical thinking, but rather scientific validation. There is a new movement in clinical medicine, which is a rational response to these demands. It is called evidence-based medicine (EBM).

Evidence-based medicine can be defined simply as the incorporation of the highest quality of information, derived from scientific study, into the context of clinical care. According to its doctrine, each practice, be it a surgical or medical intervention, should be based on research that applies the principles of the scientific method. Practically, evidence-based medicine involves defining an evidence-based problem, identifying and critically appraising the best available scientific evidence to answer the problem, and deciding on the most appropriate course of action for a given patient.

The process of patient interaction can be divided as follows: obtaining historic information, examining the patient, obtaining relevant diagnostic studies, and offering therapies. Each of these processes can lend themselves to an evidence-based question.

Back to Top | Article Outline

Evidence-based Questions Regarding the History or Physical Examination

Much of our medical training is to learn which symptoms and signs are associated with various diseases. As such, symptoms and signs can themselves be thought of as diagnostic “tests,” in that their presence or absence alters the probability for given disease. During our medical training, we learned that if vitreous cells were present in the setting of acute posterior vitreous detachment, a retinal tear must be ruled out. We recently evaluated the scientific basis for this medical truism. From our series of patients with acute retinal detachment, 13% had a retinal tear. 1 Patients who had either vitreous hemorrhage or pigmented vitreous cells were 52 times more likely to have a retinal tear. 1 The specificity of “tobacco dust” for a retinal tear was 100%, meaning that all patients with this clinical finding had a retinal tear. Accordingly, the presence of tobacco dust allows us to expect the disease of a retinal tear, and accordingly, all patients with this finding should be assumed to have a tear until proven otherwise.

Back to Top | Article Outline

Diagnostic Studies

After the history and clinical examination, we may be in a situation where we are unsure of a given diagnosis. Diagnostic studies can be employed to alter the likelihood of a given patient having a specific disease. An evidence-based question regarding a diagnostic study would be to determine the probability of a disease (posttest probability) given the results of a diagnostic study. For instance, the accuracy of autofluorescence can be determined for correctly classifying patients with adult vitelliform foveomacular dystrophy (AVFMD) from those who do not have the disease (unpublished data, S. Sharma, 2000). The presence of autofluorescence in this situation is associated with a likelihood ratio of 19. This means that patients with AVFMD are 19 times more likely to demonstrate autofluorescence than those without the disease. Knowing this likelihood ratio, if we are presented with a patient and are unsure of whether he or she has AVFMD, we can calculate their likelihood ratio for having the disease if they have foveal autofluorescence.

Let us assume that a hypothetical patient's pretest probability is 50% (i.e., the patient has a 50/50 chance of having AVFMD), and that the patient does have foveal autofluorescence. The following steps are taken to calculate the likelihood of a diagnosis of the disease:

  • Step 1: Convert the pretest probability to pretest odds.
  • Pretest odds = pretest probability / 1 − pretest probability
  • Pretest odds for this patient = 0.5 / 1.0 − 0.5 = 1.0
  • Step 2: Determine posttest odds
  • Posttest odds = (pretest odds)/positive likelihood ratio
  • Posttest odds for this patient = (1.0) × (19) = 19
  • Step 3: Converting posttest odds to posttest probability
  • Posttest probability = posttest odds / 1 + posttest odds
  • Posttest probability for this patient = 19/19 + 1 = 19 / 20 or 95%.

We have gone from a situation where we did not know whether the patient had AVFMD (50/50 chance) to one where we are very confident that he or she has the disease (95% probability).

Back to Top | Article Outline

Decision to Offer Therapy

Clinical efficacy can be measured in either absolute (absolute risk reduction) or relative terms (relative risk reduction). Efficacy as quantified in relative terms can sometime be deceiving. By reviewing the data supplied by the investigators of the Branch Retinal Vein Occlusion study, we know that 62% of patients who receive grid laser photocoagulation developed a two-line visual improvement, as compared to only 34% of those who receive no treatment. 2 Therefore, in relative terms, laser confers a 50% increase in the probability of a two-line visual improvement. Another treatment that confers a 50% relative improvement is carotid endartectomy for asymptomatic hemodynamically significant carotid artery stenosis. 3 However, in absolute terms, the endarterectomy trial demonstrated that treatment reduced the probability of a negative outcome from 2% to 1%. The absolute risk reduction associated with treatment in the Branch Retinal Vein Occlusion study was 33%, and 1% in the Asymptomatic Carotid Atherosclerosis Study.

The numbers needed to treat equals the inverse of the absolute risk reduction. It refers to the number of patients who need to be treated to obtain one treatment success. The numbers needed to treat for patients with branch retinal occlusion and macular edema is 3. This means that a clinician, hypothetically, needs to treat three patients for one patient to obtain a two-line visual improvement. The numbers needed to treat for asymptomatic patients with hemodynamically significant carotid artery stenosis is 100. This means that, theoretically, 100 patients need to have carotid endarterectomy performed in order to save one life or prevent one severe stroke.

We believe the incorporation of evidence-based principles, especially in the areas of diagnostic and therapeutic intervention, is the most effective way to practice high-quality medical care. The adoption of such practices will help insure that our patients will receive the highest quality care based on the most accurate and reproducible data available.

Back to Top | Article Outline


1. Sharma S, Walker R, Brown GC, Cruess AF. The importance of qualitative vitreous examination in patients with acute posterior vitreous detachment. Arch Ophthalmol 1999; 117:343–346.
2. Branch Retinal Vein Occlusion Study Group. Argon laser for macular edema in branch vein occlusion. Am J Ophthalmol 1984; 98:271–282.
3. Asymptomatic Carotid Atherosclerosis Study. Endarterectomy for asymptomatic carotid artery stenosis. Executive committee for the Asymptomatic Atherosclerosis Study. JAMA 1995; 273:1421–1428.
© 2002 Lippincott Williams & Wilkins, Inc.