August 16, 2004
Scientific Article Abstracts; Ephemeral Validity
M.J.McKeown, MD, FACOG, FACS
Most scientific journals require the author of an article submitted for publication to include an abstract. When one searches for articles in the National Library of Medicine in PubMed or through another source of scientific literature frequently all that is available is an abstract.
- There are many interesting definitions of what an abstractis. Some of them are as follows:
- condensation, epitome, synopsis
a statement summarizing the important points of a text
to consider without reference to a particular example
considered apart from concrete existence
a written summary of the key points of a scientific paper
However it is important to remember a quote from Sir Walter Scott; "He was incapable of forming any opinion or resolution abstracted from his own prejudices."
That quote should remind the reader of any abstract that the author wrote this short description of a larger scientific paper in such a way that it highlights what major points the author wants to make and which may not be separate from his/her strong opinions whether or not supported by statistically defensible evidence.
The previous discussions of the use of statistics included some guidelines on how to use them and how to interpret the use of them. These guidelines were to help the reader form an opinion of the validity of the data discussed and evaluate whether or not the information in the conclusions is applicable to his/her particular condition.
When an author writes an abstract it is with the definite purpose of designing it to attract a reader into reading the entire paper. This abstract may indeed include some statistical conclusions such as, "..p value less than..", "..this case control study shows that ..", or "..this meta-analysis supports the concept of..". Another interesting tip off that the author may be laying the foundation for him/her to do further studies is the inclusion of, "..the conclusions show that further work is needed in this area."
It is expected that a scientific paper will have undergone a review of its content prior to publication and indeed may have been required to re-write parts to qualify for acceptance for publication. The publication usually sends the submitted paper to one or more qualified reviewers and then an editorial board evaluates the paper and its reviews. If a paper is published it should have passed this sort of review and one should be able to trust it.
However, as discussed in the previous articles even these reviews can pass a paper whose statistics are less than perfect. Papers can be published without rigidly done statistics if they are thought to illustrate a concept that needs further review, if they are thought to express a very new concept, and, hopefully in few instances, to make a political point.
In relation to the above, it can be thought that an abstract may have little concrete value. In a way that can be true. If one is to truly evaluate a concept and is doing a literature search on a subject a review of an article abstract is likely the initial, and perhaps only, representation of an article from which to decide to search out and read the entire article. It is not unusual to browse such as the National Library of Medicine through its PubMed access and evaluate the article titles that are presented as a result of some search terms. Many of the articles will have abstracts that can be reviewed. However that is the extent of the information to be found in the results of a search. Further information requires ordering and paying for the full text of an article. If one has access to a large scientific library associated with a school or other academic institution they may order full text articles without a charge. However unless one is affiliated with the institution this is unlikely since there is a charge to the institution for this service.
The reader may now say to him/her self, "What in the world good is it to search for scientific (medical) information if its validity can't be judged within the results of the search?"
- That is an excellent question! However, since the search likely only produces an abstract at best a decision needs to be made on how much to trust that abstract and/or which few to spend money to get the full article. There is no absolutely best way to make this decision. However a few decision point suggestions are:
- It is likely that articles in scientific journals thought to be a standard of the discipline are well reviewed.
It is likely that if many articles in such publications agree then the conclusions are more to be trusted.
Money for full text articles is likely best spent on articles doing an apparently valid statistical review of many articles on a major concept.
A final suggestion would be to obtain interesting abstracts and perhaps some full text articles and then ask a trusted expert in the field to review them with you.
Remember, current knowledge, is just that and may change tomorrow, next week or next year. This means that for an individual to remain current he/she needs to continue to acquire information on the subject of interest and to continue to evaluate it themselves for evidence of major, well evaluated conclusions or increasing evidence that a conclusion is at best an opinionated observation.