Analysis of review articles in medicine
An analysis of the literature devoted even to a narrow subject, when there are many important and fulfilling high methodological criteria of articles, can be difficult. An alternative is to use information collected in review articles, which, however, to provide a source of reliable knowledge, should meet a number of criteria.
Just as incorrect test methods can lead to unjustified conclusions, including false results in a literature review can lead to bad conclusions and useless or even harmful clinical recommendations. Review articles that sum up scientific evidence, nothing can be confused with illustrative publications – descriptive in my nature in which scientific evidence and subjective opinions are mixed up.
According to the EBM strategy, review articles about important, practical clinical aspects (aetiology, diagnosis, prognosis and therapy) should clearly formulate the basic question (questions) to which the publication is to answer If it is not obvious – it is probably not worth analyzing anything.
The article, which also includes considerations on etiology, diagnosis, prognosis, treatment and prevention, is only useful for people who want to get a general idea on some subject. The conclusions appearing in such articles are not based on evidence; after declarative statements there are only references to the selected literature.
In order to assess the significance of the conclusions presented in this type of article – the reader is forced to study the quoted, original literature.
The next necessary condition is to specify what the review was about (eg: exposure, diagnostic test, prognostic symptom, treatment method, etc.), which clinical population was referred to and what result was analyzed.
To assess the effect of lowering blood cholesterol levels on the development of ischemic heart disease, the literature search may be narrowed down to the study of patients who were initially asymptomatic (primary prevention) or patients with symptoms of ischemic disease (secondary prevention). Clinical trials may only concern the use of a diet or medication, as well as both linking factors. The assessed effect may be the occurrence of ischemic disease, mortality due to ischemic disease or general mortality, regardless of the causes.
A review of the literature will only make sense if it is clearly determined which population is subject to examination (what kind of patients), the type of intervention (what prevention) and the effect being assessed.
The publication must describe the method used to perform the source literature review. Locating and reaching the source literature, even when the subject of the search has been narrowed down, often causes many problems. In the review work, a description of how to search for relevant publications is required, guaranteeing that the applied method has exhausted the available options.
The optimal situation is when the author searched:
- one or more bibliographic databases (with the selection of keywords used for the search and methods of their application),
- sets of classic articles for a given topic, found eg in the Science Citation Index,
- references attached to relevant publications, in search of citations of other important articles,
- he contacted the authors of important publications or institutions specialized in a given subject – in order to make sure. with no important report, nothing escaped attention and to locate important data, nothing published or available only in non-indexed materials.
Such a completed bibliography contains relevant articles, but also nothing meaningful. The reader should be informed how the latter have been eliminated. When the process of “screening” is not systemic, errors are created; the selection of articles becomes “selective”, which consequently often gives “positive” reviews, often corresponding to (a disgraceful in such work) assumption a priori.
It has been repeatedly shown that changing the inclusion criteria may lead to completely different conclusions of the review article, hence the choice of the literature for analysis should take place with a precise description of the patients, specific clinical effects, specific treatments, tests’, exposures or prognostic factors. The key element in the work project is the “quality filter”, which is a criterion for allowing the results to be analyzed. For example, in the modern review articles about the therapy, only the results of randomized clinical trials are allowed to be analyzed.
Correct conclusions of review articles are created after assessing the methodological quality of source publications included in the work. This assessment can be carried out bilaterally, giving the answer to the question: Do individual studies meet the minimum scientific criteria, so that the conclusions derived from them have an indisputable justification?