 |
Types of epidemiologic designs provide comparison groups, but the comparisons
may introduce bias that could affect the validity of the evaluation of
treatment effects. For example, in a comparison of patients undergoing a
new treatment with a historical cohort receiving the old treatment, there
may be temporal trends in health care or other factors related to the outcome. In addition, the quality of data available for historical controls may differ from the quality of data collected for patients receiving the new treatment because of different diagnostic criteria, measurement standards, completeness of records, etc. Thus, lack of standard criteria could yield spurious results.
Database studies tend to be inexpensive and allow evaluation of large numbers
of patients, but potential bias in treatment selection, outcome assessment,
and other factors make them a lower standard of evidence.
Although these types of studies improve on case studies or case series
because they offer comparison groups, problems with selection of patients
for treatment and lack of comparability of treatment groups complicate the
analysis. The randomized, controlled clinical trial is designed to reduce
such biases.
|
