This page provides information on the main type of clinical research biases and the research design aspects for minimising those biases.
The three main sources of bias in research studies are:
For more details about research bias, visit Catalogue of Bias, published and managed by the Centre for Evidence-Based Medicine, University of Oxford, to learn more.
To assess how serious is the risk of bias of a research study, the following 7 research design aspects and the statistical significance should be taken into account.
|
Randomisation The assignment of patients to either group (treatment or control) must be allocated randomly to eliminate selection bias. In a truly randomised design setting, all eligible participants in the study have an equal chance of being selected for either group. Methods for achieving randomisation can include a fair coin toss or computer software for generating the random allocation. Allocating patients to either group by their surnames in alphabetical order or by the date of their birthday is NOT considered as randomisation. |
Concealed allocation Concealed allocation means the sequence of assigning patients into a study arm must be concealed from the researchers. As a result, the researchers CANNOT predict or change the randomised assignment so as to further eliminate conscious or unconscious selection bias of the researchers. |
|
Similar baseline Characteristics The treatment and the control group must be similar for all prognostic characteristics except whether or not they receive the experimental treatment. |
Blinding Blinding means that everyone involved in the study, including the participants to be studied, the researchers, the data collectors / analysts and any clinicians involved in the study do not know which treatments are being given to which groups. This minimises performance and detection bias. A single blind study means the participants / patients do not know which group they have been assigned to. A double blind study means both the participants / patients and experimenters do not know who are in placebo group and who are in the intervention group. |
|
Follow-up Ideally, the number of participants in each study arm should be equal or nearly equal from the beginning to the end of the study. However patients may drop out for various reasons. Good studies are considered to have a follow-up rate of 80% or more. If there is a large number of participants who drop-out and a resulting lack of balance between groups, the researchers should explain and justify the potentials bias and limitations of the study. The planned schedule / duration of research study should also be long enough for the outcomes of interest to have a reasonable chance to occur. |
Stopped early To minimise the over-estimation of treatment / intervention effects, trials are expected to be run as planned. However, it is possible that a trial could be stopped early because of the discovery of perceived harm, or for other ethical or financial reasons etc. If a trial was run only for long enough to observe the beginning of the outcomes of interest and then stopped earlier than planned, there is considered to be a risk of overestimation of the treatment / intervention effect. |
|
" Intention-to-Treat " Analysis Once participants have been randomally allocated to a group, the participant must be analysed in the originally assigned group regardless of what actual therapy the patients took. This is called "intention-to-treat" analysis. The purpose of doing this is to maintain the randomisation so as to equalise the prognostic factors of both groups. |
Confidence Intervals (CI) + other statistical measurements "A 95% confidence interval (CI) of the mean is a range with an upper and lower number calculated from a sample. Because the true population mean is unknown, this range describes possible values that the mean could be. If multiple samples were drawn from the same population and a 95% CI calculated for each sample, we would expect the population mean to be found within 95% of these CIs. CIs are sensitive to variability in the population (spread of values) and sample size. When used to compare the means of two or more treatment groups, a CI shows the magnitude of difference between groups. This is helpful in understanding both the statistical significance and the clinical significance of a treatment" (O'Brien & Yi, 2016, p. 1680). Reference: O'Brien, S. F. & Yi, Q. L. (2016). How do I interpret a confidence interval?. Transfusion, 56(7), 1680-1683. All formulas and terms about calculations you need to know - BMJ Best Practice's Evidence based medicine (EBM) toolkit:
Key to statistical result interpretation: P-value in plain English explained from Students 4 Best Evidence, Cochrane EBM calculators from the Evidence-Based Medicine Toolbox of the Knowledge Translation Program |
|
|
© Eastern Health Library Service 2021-2023
