Evidence based medicine (EBM) is a common term we hear these days. Obviously, everyone wants to be treated based on the best information available at any given time. Researchers and statisticians commonly refer to quality of evidence and applicability of research results. Both are very closely connected to well thought out research methods, sample size and sample quality.
In an ideal situation, data from the entire human population would be collected in a perfectly designed research study. Conclusions from such study would be applicable to the entire population. But, even in this perfect case scenario, applicability to an individual is not guaranteed. There are outliers in any given sample, so what may be good for the majority of population may not be good for a particular individual.
Of course, there is no such thing as a perfect research study. Any given result of studies based on study populations with n=10, n=100, n=1000, or n=10,000 has different degree of reliability. Research methods, differences in sample population from individual groups of humans and simple typos affect the quality of what is available to us as “EBM.”
It is not uncommon for writers of scientific and clinical reviews to add citations to other reviews, which can easily lead to misquotes or “misunderstandings.” Although rare, even citation to other people’s original work can be attached to points in newer articles that claim to support completely opposite to the original finding. When someone else cites the newer article the confusion gets propagated.
Review articles commonly make statements like “research by so and so demonstrated the following result.” The information describing study sample size, methods, equipment used, statistical methods, strengths and limitations of the study is usually scarce in such statements. On occasion, the exact details are not even available in the original publications.
At minimum, we should strive to learn the following about EBM original imaging studies:
- What was the research question or hypothesis?
- What was the sample size?
- What imaging equipment was used? How precise was the imaging information (slice thickness, conventional or turbo spin echo, field of view, other details).
- Who interpreted the images and how many observers were used?
- What was considered the gold standard (ground truth)?
- What statistics were used to analyze result.
- What was the conclusion?
In this subsection, we explore what is out there in literature in relations to various topics. The details of various studies are summarized in table or outline format and very limited discussions are included. One day, natural language processing AI will be able to achieve the same goal quicker and possibly better, but more on this in another article.