171 Ashley Ave.
Charleston, SC 29425
843-792-1414
|
 |
October 2006
Evidence-Based Tip
Critical Appraisal
Last month we learned how to search for the evidence using PubMed. This month we will focus on evaluating the evidence. Although a randomized
controlled trial is the one of the best forms of evidence available, it must be a well-designed study, free from bias, in order to justify its
recommendations for patient care. We want to consider if the results of the study are valid, using the set of questions listed below.
Were the patients randomized?
How were the groups receiving the treatment intervention and the treatment comparison formed? Patient or physician preference must not determine
the arm of the trial to which the patient is assigned. Although not foolproof, randomization helps assure that the groups are as similar as possible
for factors such as severity of the disease, age, race, gender, and any other factor that might affect the results.
Was randomization concealed?
The person who is making the decision to accept a patient to a study should not be able to influence the assignment of the patient to a particular
treatment arm. Ideally, once the patient is accepted, the patient is assigned via remote randomization.
Were patients analyzed in the groups to which they were randomized?
Patient outcomes must be analyzed based on the treatment they were assigned, rather than the treatment actually received. For example, in the general
population, compliance with drug regiments can be a problem. It would exaggerate the benefit of a drug if the patients who did not take the treatment
as prescribed were not counted.
Were patients in the treatment and control groups similar with respect to know prognostic factors?
The smaller the sample size, the more difficult is it to create truly randomized groups. This is a particular danger in pediatrics, where total enrollment
is usually low. Investigators should check to be sure that the treatment intervention and comparison arms have equal distribution of factors known to be beneficial or harmful.
Were patients, clinicians, and outcome assessors aware of the group allocation?
To protect against placebo effect in patients, and to prevent caregivers from consciously or unconsciously treating or interacting with patients differently,
these groups must be blinded to group allocation. Blinding is also crucial for anyone collecting data, measuring outcomes, or analyzing data.
Was follow-up complete?
Ideally, ever person in the study would be followed for the length of time selected for the study. However, some patients' results might not be known at the
end of the study (loss to follow-up). Too many people lost will jeopardize the validity of the trial results.
There are statistical measures that can help evaluate the study as well. The final step in the patient-centered, evidence-based approach is to ensure the results
of the data can be applied to our patient.
|
|
Print Version